FOR PUBLICATION

UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

REYNALDO GONZALEZ; THE No. 18-16700 ESTATE OF NOHEMI GONZALEZ; BEATRIZ GONZALEZ, Individually D.C. No. and as Administrator of the Estate 4:16-cv-03282- of Nohemi Gonzalez; JOSE DMR HERNANDEZ; REY GONZALEZ; PAUL GONZALEZ, Plaintiffs-Appellants, v. GOOGLE LLC, Defendant-Appellee.

Appeal from the United States District Court for the Northern District of California Donna M. Ryu, Magistrate Judge, Presiding

2 GONZALEZ V. GOOGLE

MEHIER TAAMNEH; LAWRENCE No. 18-17192 TAAMNEH; SARA TAAMNEH; DIMANA TAAMNEH, D.C. No. Plaintiffs-Appellants, 3:17-cv-04107- EMC v.

TWITTER, INC.; GOOGLE LLC; FACEBOOK, INC., Defendants-Appellees.

Appeal from the United States District Court for the Northern District of California Edward M. Chen, District Judge, Presiding

GONZALEZ V. GOOGLE 3

GREGORY CLAYBORN, Individually No. 19-15043 and as Successor-In-Interest of the Estate of SIERRA CLAYBORN; KIM D.C. Nos. CLAYBORN; TAMISHIA CLAYBORN; 3:17-cv-06894-LB VANESSA NGUYEN, Individually 3:18-cv-00543-LB and as Successor-In-Interest of the Estate of TIN NGUYEN; TRUNG DO; JACOB THALASINOS; JAMES OPINION THALASINOS, Plaintiffs-Appellants, v. TWITTER, INC.; FACEBOOK, INC.; GOOGLE LLC, Defendants-Appellees.

Appeal from the United States District Court for the Northern District of California Laurel D. Beeler, Magistrate Judge, Presiding

Argued and Submitted March 26, 2020 San Francisco, California

Filed June 22, 2021

Before: Ronald M. Gould, Marsha S. Berzon, and Morgan Christen, Circuit Judges.

Opinion by Judge Christen; Concurrence by Judge Berzon; Partial Concurrence and Partial Dissent by Judge Gould

4 GONZALEZ V. GOOGLE

SUMMARY*

Anti-Terrorism Act

The panel addressed appeals from the district court’s dismissal of three actions seeking damages under the Anti- Terrorism Act against Google, Twitter, and Facebook on the basis that defendants’ social media platforms allowed ISIS to post videos and other content to communicate the terrorist group’s message, to radicalize new recruits, and to generally further its mission. The panel affirmed the judgments in the Gonzalez and Clayborn appeals and reversed and remanded in Taamneh.

Members of the families of victims of terrorism in Paris, Istanbul, and San Bernardino alleged that Google, Twitter, and Facebook were directly and secondarily liable for ISIS’s acts of international terrorism. The Gonzalez plaintiffs brought claims for both direct and secondary liability against Google. The district court concluded that most of plaintiffs’ claims were barred pursuant to 47 U.S.C. § 230 of the Communications Decency Act, and the direct liability claims failed to adequately allege proximate cause. In the Taamneh and Clayborn cases, the district court concluded that plaintiffs failed to plausibly allege a secondary liability claim against Google, Twitter, and Facebook.

The panel held that the district court in Gonzalez properly ruled that § 230 barred most of plaintiffs’ claims. The panel further held that the Gonzalez plaintiffs failed to state an

* This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.

GONZALEZ V. GOOGLE 5

actionable claim as to their remaining theories of liability. In Taamneh, the panel held that the district court erred by ruling that plaintiffs failed to state a claim for aiding-and-abetting liability under the ATA. In Clayborn, the panel concluded that the district court correctly held that plaintiffs failed to plausibly plead their claim for aiding-and-abetting liability.

Addressing Gonzalez, the panel held that the civil remedies section of the ATA permits United States nationals to recover damages for injuries suffered “by reason of acts of international terrorism.” The Justice Against Sponsors of International Terrorism Act of 2016 (JASTA) amended the ATA to include secondary civil liability for aiding and abetting, or conspiring to commit, acts of international terrorism. Section 230 of the Communications Decency Act protects websites from liability for material posted on the website by someone else. The panel held that the presumption against the extraterritorial application of federal statutes did not prevent § 230 from applying to the Gonzalez plaintiffs’ claims because the relevant conduct took place in the United States. The panel concluded that JASTA did not impliedly repeal § 230. Agreeing with the First and Second Circuits, the panel held that the exception set forth in § 230(e)(1), concerning impairment of the enforcement of federal criminal statutes, does not extend to actions for civil damages. Thus, the Gonzalez plaintiffs’ claims were not categorically excluded from the reach of § 230 immunity.

The Gonzalez plaintiffs argued that the immunity afforded by § 230 did not bar their claims because § 230 immunizes only those who publish content created by third parties, and their claims were directed to content created by Google. Google argued that the plaintiffs impermissibly sought to treat Google as a publisher of content created by third parties,

6 GONZALEZ V. GOOGLE

presumably ISIS, on YouTube. In Part III.E of its opinion, the panel affirmed the district court’s ruling that § 230 barred all of plaintiffs’ claims except to the extent their complaint presented claims premised on the allegation that Google shared advertising revenue with ISIS. Section 230(c) precludes liability for “(1) a provider or user of an interactive computer service (2) whom a plaintiff seeks to treat . . . as a publisher or speaker (3) of information provided by another information content provider.” The panel concluded that plaintiffs’ claims did not inherently require the court to treat Google as the publisher or speaker of content provided by ISIS, and the duty that plaintiffs alleged Google violated did not derive from Google’s status or conduct as a publisher or speaker. The panel concluded that Google did not create or develop content by making a material contribution to its alleged unlawfulness when it created the “mosaics” by which ISIS videos were delivered. The panel held that the court’s case law foreclosed the argument that Google’s pairing of ISIS content with selected advertising and other videos vitiated § 230 immunity. Accepting as true plaintiffs’ allegation that Google’s algorithms recommended ISIS content to users, and agreeing with the Second Circuit, the panel wrote that the algorithms did not treat ISIS-created content differently than any other third-party created content, and thus were entitled to § 230 immunity.

In Part III.F of its opinion, the panel held that § 230 did not bar the Gonzalez plaintiffs’ claims premised on the allegation that because it shared advertising revenue with ISIS, Google should be held directly liable for providing material support to ISIS and secondarily liable for providing substantial assistance to ISIS.

GONZALEZ V. GOOGLE 7

In Parts IV and V, the panel held that the Gonzalez plaintiffs did not adequately allege claims for direct or secondary liability under the ATA based on a revenue-sharing theory. As to direct liability, plaintiffs failed to plausibly allege that Google directly perpetrated an act of international terrorism because they did not allege that Google’s actions were motivated by anything other than economic self- enrichment. As to secondary liability, plaintiffs did not state a claim on either a theory of aiding and abetting or a theory of conspiracy liability.

In Part VI, reversing the district court’s dismissal of the Taamneh action, the panel held that the Taamneh plaintiffs adequately stated a claim for aiding-and-abetting liability.

In Part VII, affirming the dismissal of the Clayborn action, the panel held that because the Clayborn plaintiffs did not plausibly allege that ISIS committed, planned or authorized the terrorist attack in San Bernardino, they did not adequately state a claim for aiding and abetting an act of international terrorism.

Judge Berzon concurred in the majority opinion in full. She wrote separately to explain that, although the panel was bound by Ninth Circuit precedent compelling the outcome in this case, she joined the growing chorus of voices calling for a more limited reading of the scope of § 230 immunity. Judge Berzon urged the court to reconsider its precedent en banc to the extent that it holds that § 230 immunity extends to the use of machine-learning algorithms to recommend content and connections to users.

Judge Gould concurred in the majority opinion in its Parts I and II, Part III.A through III.D, Part III.F, and Part VI and

8 GONZALEZ V. GOOGLE

dissented in part as to Part III.E and Parts IV, V, and VII. Judge Gould wrote that he concurred insofar as the majority would reverse in part the dismissal of revenue-sharing claims in Gonzalez, and insofar as it would reverse the district court’s judgment in Taamneh that the complaint failed to adequately state a claim under the ATA. Judge Gould wrote that he dissented as to the majority’s dismissal of the Gonzalez claims on grounds of § 230 immunity, and of failure to state a claim for direct or secondary liability under the ATA, because of the majority’s mistaken conclusion that there was no act of international terrorism, and he also would hold that the complaint adequately alleged that there was proximate cause supporting damages on those claims. Judge Gould agreed that claims could proceed in the Taamneh case, and accordingly agreed with reversing and remanding in that case. On the Clayborn case, Judge Gould dissented because the majority’s conception of an attack authorized by ISIS was inconsistent with the allegations of the operative complaint and well-established principles of tort and agency law.

GONZALEZ V. GOOGLE 9

COUNSEL

Keith Altman (argued) and Daniel W. Weininger (argued), Excolo Law, Southfield, Michigan, Plaintiffs-Appellants Reynaldo Gonzalez, Mehier Taamneh, Lawrence Taamneh, Sara Taamneh, Dimana Taamneh, Gregory Clayborn, Kim Clayborn, Tamishia Clayborn, Vanessa Nguyen, Trung Do, Jacob Thalasinos, and James Thalasinos.

Robert J. Tolchin (argued) and Meir Katz, Berkman Law Office LLC, Brooklyn, New York; for Plaintiffs-Appellants Estate of Nohemi Gonzalez; Beatriz Gonzalez, Jose Hernandez, Rey Gonzalez, and Paul Gonzalez.

Brian M. Willen (argued), Wilson Sonsini Goodrich & Rosati, New York, New York; David H. Kramer, Lauren Gallo White, and Kelly M. Knoll, Wilson Sonsini Goodrich & Rosati, Palo Alto, California; for Defendant-Appellee Google LLC.

Kristin A. Linsley (argued) and Jacob T. Spencer, Gibson Dunn & Crutcher LLP, San Francisco, California; for Defendant-Appellee Facebook Inc.

Seth P. Waxman, Patrick J. Carome, and Ari Holtzblatt, Wilmer Cutler Pickering Hale & Dorr LLP, Washington, D.C., for Defendant-Appellee Twitter Inc.

Aaron Mackey and Sophia Cope, Electronic Frontier Foundation, San Francisco, California, for Amicus Curiae Electronic Frontier Foundation.

10 GONZALEZ V. GOOGLE

OPINION

CHRISTEN, Circuit Judge:

We address three appeals arising from separate acts of terrorism—one in Paris, one in Istanbul, and one in San Bernardino—in which Nohemi Gonzalez, Nawras Alassaf, Sierra Clayborn, Tin Nguyen, and Nicholas Thalasinos lost their lives. The foreign terrorist organization known as ISIS took responsibility for the attacks in Paris and Istanbul and lauded the attack in San Bernardino after the fact. Plaintiffs are members of the victims’ families.

Plaintiffs seek damages pursuant to the Anti-Terrorism Act (ATA), 18 U.S.C. § 2333. The ATA allows United States nationals to recover damages for injuries suffered “by reason of an act of international terrorism,” id. § 2333(a), but the defendant in these cases is not ISIS. Instead, plaintiffs allege that Google, Twitter, and Facebook are directly and secondarily liable for the five murders at issue in these cases. The complaints allege that defendants’ social media platforms allowed ISIS to post videos and other content to communicate the terrorist group’s message, to radicalize new recruits, and to generally further its mission. Plaintiffs also claim that Google placed paid advertisements in proximity to ISIS-created content and shared the resulting ad revenue with ISIS. In these and other ways, all three complaints allege defendants are directly liable for committing acts of international terrorism pursuant § 2333(a) of the ATA, and secondarily liable for conspiring with, and aiding and

GONZALEZ V. GOOGLE 11 abetting, ISIS’s acts of international terrorism pursuant to § 2333(d).1

This opinion addresses three separate appeals. The Gonzalez appeal concerns claims for both direct and secondary liability against Google. In that case, the district court granted Google’s motion to dismiss, concluding that most of the Gonzalez Plaintiffs’ claims were barred pursuant to 47 U.S.C. § 230 of the Communications Decency Act (CDA), and that the Gonzalez Plaintiffs’ direct liability claims failed to adequately allege proximate cause. The Taamneh and Clayborn appeals concern claims for secondary liability against Google, Twitter, and Facebook. In both of these cases, the district court granted defendants’ motions to dismiss on the grounds that the plaintiffs failed to plausibly allege a secondary liability claim under the ATA.

We have jurisdiction pursuant to 28 U.S.C. § 1291. We conclude the district court in Gonzalez properly ruled that § 230 bars most of the Gonzalez Plaintiffs’ claims, and that the Gonzalez Plaintiffs failed to state an actionable claim as to their remaining theories of liability asserted pursuant to the ATA. In Taamneh, we conclude the district court erred by ruling the Plaintiffs failed to state a claim for aiding-and- abetting liability under the ATA. The district court did not reach § 230 immunity in Taamneh. In Clayborn, we conclude the district court correctly held that Plaintiffs failed As to the first element of § 230, the parties do not dispute that Google is an “interactive computer service” provider as defined in 47 U.S.C. § 230(f)(2). We agree. Roommates, 521 F.3d at 1162 n.6 (“[T]he most common interactive computer services are websites.”); see also Kimzey v. Yelp!, Inc., 836 F.3d 1263, 1268 (9th Cir. 2016) (“Yelp is plainly a provider of an ‘interactive computer service’ . . . , a term that

the basis for dismissal for failure to state a claim depends on whether the allegations in the complaint suffice to establish that ground.” Jones v. Bock, 549 U.S. 199, 215 (2007). Here, “the ‘allegations in the complaint suffice to establish’ the defense,” and thus the “affirmative defense may be considered properly.” Sams v. Yahoo! Inc., 713 F.3d 1175, 1179 (9th Cir. 2013) (quoting Jones, 549 U.S. at 215); see also Ricci v. Teamsters Union Local 456, 781 F.3d 25, 27 (2d Cir. 2015) (per curiam) (considering whether § 230 immunity barred plaintiffs’ claims on a 12(b)(6) motion to dismiss). 9 Barnes limited its summary of § 230(c)(1) eligibility requirements to instances where “plaintiff[s] seeks to treat [the defendant], under a state law cause of action, as a publisher or speaker” because that case only concerned state law claims. 570 F.3d at 1100 (emphasis added). In Roommates, we acknowledged that § 230 immunity is not limited to cases in which plaintiffs assert state law claims. 521 F.3d at 1164; see also Barnes, 570 F.3d at 1100 n.4.

34 GONZALEZ V. GOOGLE we interpret expansively under the CDA.” (quotations and alterations omitted)).

As to the second element, the Gonzalez Plaintiffs argue their claims do not inherently require a court to treat Google as a publisher or speaker. Google responds that the thrust of the Gonzalez Plaintiffs’ claims is that Google did not do enough to block or remove content, and that such claims necessarily require the court to treat Google as a publisher. On this point, we agree with Google.

What matters when we assess this element is “whether the cause of action inherently requires the court to treat the defendant as the ‘publisher or speaker’ of content provided by another.” Barnes, 570 F.3d at 1102. This element is satisfied when “the duty that the plaintiff alleges the defendant violated derives from the defendant’s status or conduct as a ‘publisher or speaker.’” Id.

The Gonzalez Plaintiffs argue that their claims do not treat Google as a publisher, but instead assert a simple “duty not to support terrorists.” They maintain that just as the ATA prohibits a retailer like Wal-Mart “from supplying fertilizer, knives, or even food to ISIS,” the ATA prohibits Google from supplying ISIS with a communication platform. The Gonzalez Plaintiffs’ characterization of their claim as asserting a “duty not to support terrorists” overlooks that publication itself is the form of support Google allegedly provided to ISIS. See Force, 934 F.3d at 65 (recognizing that supplying a platform and communication services “falls within the heartland of what it means to be the ‘publisher’ of information under Section 230(c)(1)”). The Plaintiffs’ non-

GONZALEZ V. GOOGLE 35 revenue sharing claims seek to impose liability for the content Google allowed to be posted on its platform.

Publishing encompasses “any activity that can be boiled down to deciding whether to exclude material that third parties seek to post online . . . .” Roommates, 521 F.3d at 1170–71. “[P]ublication involves reviewing, editing, and deciding whether to publish or to withdraw from publication third-party content.” Barnes, 570 F.3d at 1102; see also Klayman v. Zuckerberg, 753 F.3d 1354, 1359 (D.C. Cir. 2014) (“[T]he very essence of publishing is making the decision whether to print or retract a given piece of content . . . .”). Here, the Gonzalez Plaintiffs assert that Google failed to prevent ISIS from using its platform, and thereby allowed ISIS to disseminate its message of terror. Because the non- revenue sharing claims seek to impose liability for allowing ISIS to place content on the YouTube platform, they seek to treat Google as a publisher.

The Gonzalez Plaintiffs argue that Google does more than merely republish content created by third parties; the TAC alleges that Google “creat[es]” and “develop[s]” the ISIS content that appears on YouTube, at least in part, and therefore receives no protection under § 230. Again, we disagree. This argument is precluded by this court’s § 230 precedents.

The Gonzalez Plaintiffs are correct that § 230 immunity only applies to the extent interactive computer service providers do not also provide the challenged information content. Roommates, 521 F.3d at 1162–63; see also Carafano, 339 F.3d at 1123. An “information content

36 GONZALEZ V. GOOGLE

provider” is defined as “any person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service.” 47 U.S.C. § 230(f)(3) (emphasis added).

We have held that a website that “creat[es] or develop[s]” content “by making a material contribution to [its] creation or development” loses § 230 immunity. Kimzey, 836 F.3d at 1269. A “material contribution” does not refer to “merely . . . augmenting the content generally, but to materially contributing to its alleged unlawfulness.” Roommates, 521 F.3d at 1167–68 (emphasis added). This test “draw[s] the line at the ‘crucial distinction between, on the one hand, taking actions” to display “actionable content and, on the other hand, responsibility for what makes the displayed content [itself] illegal or actionable.” Kimzey, 836 F.3d at 1269 n.4 (internal quotation marks omitted) (quoting Jones v. Dirty World Ent. Recordings LLC, 755 F.3d 398, 413–14 (6th Cir. 2014)). Other circuits have adopted this “material contribution” test, acknowledging that making a material contribution does not mean “merely taking action that is necessary to the display of the allegedly illegal content,” but rather, “being responsible for what makes the displayed content allegedly unlawful.” Dirty World Ent., 755 F.3d at 410; see also, e.g., FTC v. LeadClick Media, LLC, 838 F.3d 158, 176 (2d Cir. 2016); Klayman, 753 F.3d at 1358; Nemet Chevrolet, Ltd. v. Consumeraffairs.com, Inc., 591 F.3d 250, 257–58 (4th Cir. 2009); FTC v. Accusearch Inc., 570 F.3d 1187, 1197–1201 (10th Cir. 2009). Absent this sort of “material contribution,” Google does not qualify as an

GONZALEZ V. GOOGLE 37

“information content provider,” and may be eligible for § 230 immunity. See Kimzey, 836 F.3d at 1269–70.10

Plainly, an interactive computer service does not create or develop content by merely providing the public with access to its platform. A “website does not create or develop content when it merely provides a neutral means by which third parties can post information of their own independent choosing online.” Kimzey, 836 F.3d at 1270 (quoting Klayman, 753 F.3d at 1358). Thus, in Kimzey, we concluded that a provider does not create or develop content when its website “does ‘absolutely nothing to enhance the defamatory sting of the message’ beyond the words offered by the [third- party] user.” Id. (quoting Roommates, 521 F.3d at 1172).

The Gonzalez Plaintiffs concede that Google did not initially create any ISIS videos, but allege that Google creates the “mosaics” by which that content is delivered. According to the Gonzalez TAC, Google makes a material contribution to the unlawfulness of ISIS content by pairing it with selected advertising and other videos because “pairing” enhances user engagement with the underlying content. Our case law forecloses the argument that this type of pairing vitiates § 230 immunity.

The first element of aiding and abetting liability requires a showing that the party the defendant aided committed an act of international terrorism that injured the plaintiff. ^[18] U.S.C. § 2333(d)(2); Halberstam, 705 F.2d at 477; see also Siegel, 933 F.3d at 223.16 The parties dispute whether the relevant principal actor is the ISIS organization as a whole or the individual terrorists who perpetrated the Paris Attacks. We agree with the Gonzalez Plaintiffs that ISIS is the relevant The second element of aiding-and-abetting liability requires a showing that Google was generally aware of its role in ISIS’s terrorist activities at the time it provided assistance to ISIS. ^[18] U.S.C. § 2333(d)(2); Halberstam,

58 GONZALEZ V. GOOGLE

705 F.2d at 477; see also Linde, 882 F.3d at 329. The Gonzalez Plaintiffs also satisfied this element.

Just as the Halberstam court concluded that Linda Hamilton was generally aware of her role in Bernard Welch’s ongoing burglary operation because she “knew about and acted to support” it, the Gonzalez Plaintiffs must plausibly allege that, by sharing revenue with ISIS, Google was aware that it was assuming a role in ISIS’s terrorist activities. See Halberstam, 705 F.2d at 488; see also Linde, 882 F.3d at 329 (requiring a showing that “the bank was ‘generally aware’ that [by providing financial services,] it was thereby playing a ‘role’ in Hamas’s violent or life-endangering activities” (quoting Halberstam, 705 F.2d at 477)). Notably, this element does not require a showing of “the specific intent demanded for criminal aiding and abetting culpability,” i.e., an “intent to participate in a criminal scheme as ‘something that he wishes to bring about and seek by his action to make it succeed.’” Linde, 882 F.3d at 329 (quoting Rosemond v. United States, 572 U.S. 65, 76 (2014)). Nor does it require that Google “knew of the specific attacks at issue.” Id.

The TAC adequately alleges that Google was aware of the role it played in ISIS’s terrorist activities. Specifically, the Gonzalez Plaintiffs allege that Google knowingly shared advertising revenue with ISIS and that Google did so despite numerous reports from news organizations that Google placed advertisements on ISIS videos. Under these circumstances, the allegation that Google knowingly gave “fungible dollars to a terrorist organization” plausibly alleges that Google was aware of the role it played in activities that “may be ‘dangerous to human life.’” Cf. Kemper, 911 F.3d at 390; see also Fields, 881 F.3d at 748; Boim, 549 F.3d at 693.

GONZALEZ V. GOOGLE 59

We are mindful that “aiding and abetting an act of international terrorism requires more than the provision of material support to a designated terrorist organization.” Linde, 882 F.3d at 329. Thus, the mens rea required for the general awareness element of secondary liability under § 2333(d) may not be coextensive with the showing required for material support under § 2339B. The latter “requires only knowledge of the organization’s connection to terrorism, not intent to further its terrorist activities or awareness that one is playing a role in those activities.” See id. at 330 (citing Holder, 561 U.S. at 16–17); see also, e.g., Siegel, 933 F.3d at 224 (concluding plaintiffs failed to plead general awareness with allegations “suggest[ing] that in providing banking services to [a Saudi Arabian bank], HSBC had little reason to suspect that it was assuming a role in [al-Qaeda in Iraq’s] terrorist activities”). But here, we are satisfied that the allegations indicating Google knowingly contributed money to ISIS suffice to show that Google understood it played a role in the violent and life-endangering activities undertaken by ISIS, and therefore establish the second element of aiding- and-abetting liability for purposes of § 2333(d)(2).

The third element of aiding-and-abetting liability requires that the plaintiff show the defendant knowingly and substantially assisted the act of terrorism that injured the plaintiff. ^[18] U.S.C. § 2333(d)(2); see also Halberstam, 705 F.2d at 488 (holding the defendant must have “knowingly and substantially assist[ed] the principal violation”). This element contains two components: (1) “knowing[]” assistance, and (2) “substantial[]” assistance. See Halberstam, 705 F.2d at 477; see also id. at 488 (evaluating whether Linda Hamilton assisted Bernard Welch “with

60 GONZALEZ V. GOOGLE

knowledge that he had engaged in illegal acquisition of goods” separate from considering whether her “assistance was ‘substantial’”).

The Halberstam court identified six factors relevant to assessing whether the substantial assistance component is satisfied: “(1) the nature of the act encouraged, (2) the amount of assistance given by defendant, (3) defendant’s presence or absence at the time of the tort, (4) defendant’s relation to the principal, (5) defendant’s state of mind, and (6) the period of defendant’s assistance.” Linde, 882 F.3d at 329 (citing Halberstam, 705 F.2d at 483–84).17

The parties dispute whether the relevant “principal violation” for analyzing the third element is ISIS’s broader campaign of terrorism or the Paris Attacks. See Halberstam, 705 F.2d at 488. But Halberstam explained that the extent of liability under aiding-and-abetting encompasses foreseeability, such that a defendant “who assists a tortious act may be liable for other reasonably foreseeable acts done in connection with it.” 705 F.2d at 484. For example, the common law cases Halberstam drew upon established that a thirteen-year-old boy who broke into a church with some Taamneh Plaintiffs’ aiding-and-abetting claim is governed by the standards set forth in Halberstam. The first Halberstam element requires that “the party whom the The second Halberstam element of aiding-abetting liability requires the defendant to be “generally aware of his role as part of an overall illegal or tortious activity at the time that he provides the assistance.” Id. The Taamneh Plaintiffs also satisfied this element.

The Taamneh Plaintiffs allege that, at the time of the Reina Attack, defendants were generally aware that ISIS used defendants’ platforms to recruit, raise funds, and spread propaganda in support of their terrorist activities. The FAC alleges that, despite “extensive media coverage” and legal and governmental pressure, defendants “continued to provide these resources and services to ISIS and its affiliates, refusing to actively identify ISIS’s Twitter, Facebook, and YouTube accounts, and only reviewing accounts reported by other social media users.” These allegations suggest the defendants, after years of media coverage and legal and government pressure concerning ISIS’s use of their platforms, were generally aware they were playing an important role in ISIS’s terrorism enterprise by providing access to their platforms and not taking aggressive measures to restrict ISIS-affiliated content. See Linde, 882 F.3d at 329; see also Halberstam, 705 F.2d at 477.

70 GONZALEZ V. GOOGLE The third Halberstam element requires the plaintiff to allege the defendant knowingly and substantially assisted the principal violation. 705 F.2d at 477. We conclude the Taamneh Plaintiffs’ complaint satisfied this element.

The Taamneh Plaintiffs adequately allege that defendants knowingly assisted ISIS. Specifically, the FAC alleges that ISIS depends on Twitter, Facebook, and YouTube to recruit individuals to join ISIS, to promote its terrorist agenda, to solicit donations, to threaten and intimidate civilian populations, and to inspire violence and other terrorist activities. The Taamneh Plaintiffs’ complaint alleges that each defendant has been aware of ISIS’s use of their respective social media platforms for many years—through media reports, statements from U.S. government officials, and threatened lawsuits—but have refused to take meaningful steps to prevent that use. The FAC further alleges that Google shared revenue with ISIS by reviewing and approving ISIS’s YouTube videos for monetization through the AdSense program. Taken as true, these allegations sufficiently allege that defendants’ assistance to ISIS was knowing.

We next consider whether the Taamneh Plaintiffs plausibly allege that defendants’ assistance was “substantial,” applying the six Halberstam factors.^[21] First, the act The majority acknowledges that Section 230 does not shield Google from liability on the revenue sharing claims because the allegations are “premised on Google providing ISIS with material support by giving ISIS money.” I concur with that aspect of the opinion, but I would also add that providing monetary support to a foreign terrorist organization, with the constructive knowledge that that money would likely be used as part of the terrorist enterprise, qualifies as an “act of international terrorism.” 18 U.S.C. § 2333(a).

I begin with the contours of Plaintiffs’ revenue sharing claim. The complaint alleges that Google is aware of ISIS’s presence on YouTube because it has received complaints about ISIS content, and it has “suspended or blocked selected ISIS-related accounts at various times.” Plaintiffs also allege that Google shares a percentage of the revenue it generates from pairing advertisements and videos with the video poster. Through Google’s commercial service, AdSense, users can register their accounts for “monetization.” Plaintiffs allege that ISIS uses the AdSense monetization program to earn revenue. Before the YouTube video can be approved for advertisements, Google must review and approve the video. Google has therefore “reviewed and approved ISIS videos,

108 GONZALEZ V. GOOGLE

including videos posted by ISIS-affiliated users, for ‘monetization’ through Google’s placement of ads in connection with those videos.” Through those approvals, Google gains constructive knowledge of the fact that it provided financial support to ISIS and incentivized ISIS to continue to post videos on YouTube. Plaintiffs’ allegations about Google’s knowledge is bolstered by contentions that various news outlets reported on the kind of ads appearing before ISIS YouTube videos.

The majority mistakenly concludes that Google’s conduct could not qualify as international terrorism because it is not “intended to intimidate or coerce a civilian population or to influence or affect a government.” I disagree. The standard for intent under the ATA is not subjective; rather, it is a “matter of external appearance.” Boim v. Holy Land Found. for Relief & Dev., 549 F.3d 685, 694 (7th Cir. 2008) (en banc). I would hold that, on the facts alleged, a knowing provision of resources to a terrorist organization constitutes aid to international terrorism because an entity like Google appears to intend the natural and foreseeable consequences of its actions. See Restatement (Second) of Torts, § 8A (1965).

The majority relies on Linde v. Arab Bank, PLC, 882 F.3d 314 (2d Cir. 2018), to conclude that knowingly providing material support to a terrorist organization is not “an act of international terrorism” if it is motivated by economics. Besides the fact that Linde is a sister circuit decision that is not binding on our court, its facts and holding are also distinguishable. In Linde, the court expressly held that it was error for the district court to instruct the jury that proof that Arab Bank provided material support to a designated foreign terrorist organization, in violation of § 2339B, “necessarily proved the bank’s commission of an act of international

GONZALEZ V. GOOGLE 109

terrorism.” Id. at 325. Thus, the Second Circuit held only that violating § 2339B does not inherently create an act of terrorism. The court’s reasoning continually references the context of its decision: whether it could find that the jury instruction error was harmless. Id. at 327 (holding that “the mere provision of routine banking services to organizations and individuals said to be affiliated with terrorists does not necessarily establish causation”) (internal quotation marks and citation omitted) (emphasis added). Indeed, the court did not even decide whether Arab Bank’s financial services to Hamas should be viewed as “routine” under the court’s precedent, because that issue raised a question of fact for the jury to decide. Id.

Even accepting that providing material aid “does not invariably equate” to an act of international terrorism under § 2331(1), Linde, 883 F.3d at 326, there are clearly situations where providing such aid operates to endanger human life and manifests an apparent intent to coerce or intimidate civilians or to influence or affect governments. The Seventh Circuit’s Boim decision represents such a case, despite the majority incorrectly characterizing Plaintiffs’ reliance on it as “misplaced.” In Boim, the court held that a jury could find defendants liable under the ATA when they had donated money to Hamas and Hamas-affiliated charities, knowing that Hamas used such money to finance violence towards at least some American citizens. 549 F.3d at 690. Because donating money to Hamas was like “giving a loaded gun to a child,” it did not matter that the act of giving money is not a violent act itself because, in context, it would be “dangerous to human life.” Id. (citation omitted). The Seventh Circuit recognized that imposing liability for providing money to a terrorist group “makes good sense as a counterterrorism measure,” because “[d]amages are a less effective remedy against

110 GONZALEZ V. GOOGLE terrorists and their organizations than against their financial angels.” Id.

Boim relied on the foreseeability of the consequences of donating to Hamas to support its sensible holding that the donations would appear to be intended to intimidate or coerce a civilian population. Id. at 694; see also Linde, 883 F.3d at 327 (discussing Boim’s reasoning and stating that “given such foreseeable consequences,” the donations met the statutory definition for an act of terrorism). The court analogized donating to a terrorist organization to giving a small child a loaded gun because in both cases, the actor is “doing something extremely dangerous and without justification.” Id. at 693. “If the actor knows that the consequences are certain, or substantially certain, to result from his act, and still goes ahead, he is treated by the law as if he had in fact desired to produce the result.” Id. (quoting Restatement (Second) of Torts, § 8A (1965)). The fact that the actor was not motivated by a desire for the child to shoot anyone is of no matter to the tort inquiry. Id.

The Gonzalez Plaintiffs allege that Google knew ISIS was using its AdSense program, and that therefore Google knew it was providing material support to a terrorist organization. The fact that Google was not motivated by a desire to augment ISIS’s efforts to recruit other terrorists is irrelevant. The majority’s argument—that Google’s interactions with ISIS via revenue sharing are not intended to intimidate or coerce civilian populations because Google was “motivated by economics”—is an arbitrary line divorced from Section 2333’s text and established principles of tort law. Boim—a decision properly based upon Section 2333’s text and history—does not attempt to draw a line based on motivation. In fact, it rejects such a line as irrelevant to the question of

GONZALEZ V. GOOGLE 111

intent because a person intends what he knows is substantially certain to result from his act. 549 F.3d at 693. My colleagues attempt to distinguish Boim by noting that a donor to Hamas would likely share that organization’s vision and objectives, but Boim did not rely on that aspect of targeted donation. Instead, the Seventh Circuit reasoned that “[a] knowing donor to Hamas” is “a donor who knew the aims and activities of the organization.” Id. at 693–94 (emphasis added). It was the donor’s knowledge of Hamas’ activities, rather than his approval of it, that gave rise to liability.

Because amplifying ISIS’s message and creating new networks of prospective terrorist recruits foreseeably provides material support to a terrorist organization, I would likewise hold that the complaint in Gonzalez v. Google states a claim that Google is primarily liable on a non-revenue sharing theory.

Terrorism is, in part, psychological warfare. The record shows that for ISIS terrorism is a psychological weapon. ISIS’s most potent and far-reaching weapon is the Internet. The Gonzalez complaint alleges that “Google’s YouTube platform has played an essential role in the rise of ISIS,” which has become one of the largest perpetrators of violence in the world. ISIS uses YouTube to recruit members, plan terrorist attacks, issue threats, take credit for attacks, and demand and attempt to obtain results from the attacks by influencing government policies and conduct. While one of ISIS’s goals is to commit acts of violence, “the physical attack itself and the harm to the individual victims of the attack” is just one piece of the puzzle—ISIS also uses terror

112 GONZALEZ V. GOOGLE

attacks as a means to communicate its political message and instill fear in those it considers its combatants. Thus, the impact of ISIS’s terrorism is dependent upon its ability to communicate its message and reach its intended audiences. Id. Plaintiffs allege that “ISIS’s use of violence and threats of violence [are] part of its program of terrorism, designed . . . to gain attention, instill fear and ‘terror’ in others, send a message, and obtain results.” Because the communication of ISIS violence and threats is part of the terrorist attack, repeated postings and encouraged viewings of ISIS videos, as effected by Google’s algorithms, is also part of the attack.

When a terrorist group blows up or shoots up or carves up passengers on an airplane, railroad car or a subway car, they do not do it merely to destroy property or injure people involved in those bombings, shootings, and knifing attacks. Instead, they aim to create fear in the public so that people will be afraid to use airplanes or railroad cars or subways or any general public area to go about their business as usual. Publicizing the event is just as essential to terrorists’ success as is the bombing, shooting, or knifing itself. So-called “neutral” algorithms created by Facebook, Twitter, and Google, are then transformed into deadly missiles of destruction by ISIS, even though they were not initially intended to be used that way. But once there is a consistent stream of conduct by ISIS, it should be understood that defendants who passively ignore that conduct can be held to have intended the natural and probable consequences of their actions. See Restatement (Second) of Torts, § 8A (1965).

Just as sharing revenue with ISIS is “dangerous to human life,” Boim, 549 F.3d at 690 (citation omitted), so is amplifying its message and encouraging recruitment to its ranks. Perhaps even more so because unlike money, which

GONZALEZ V. GOOGLE 113

is fungible, YouTube has a virtual monopoly on hosting extremist videos.^[10] ISIS can get operating funds from a variety of sources, but very few platforms have the international network and infrastructure to which YouTube has access. Imposing liability on social media platforms for affirmatively amplifying ISIS’s message can therefore “cut the terrorists’ lifeline.” See id. at 691.

B

Direct liability claims under the ATA require that plaintiffs show they suffered injury “by reason of an act of international terrorism.” 18 U.S.C. § 2333(a). The “by reason of” phrasing has been understood to impose a requirement of proximate causation. See, e.g., Fields v. Twitter, 881 F.3d 739, 744 (9th Cir. 2018). To meet this requirement, “a plaintiff must show at least some direct relationship between the injuries that he or she suffered and the defendant’s acts.” Id. at 744.

On my view of the case, the proximate cause issue must be reached, and I believe that it is satisfied. The ATA’s purpose in part is to provide a financial remedy to victims of terrorism. Indeed, ATA’s legislative history demonstrates Congress’s intent to authorize the “imposition of liability at any point along the causal chain of terrorism.” S. Rep. No. 102-342, at 22 (1992) (referencing “the flow of money” to terrorist groups).

10 See, e.g., Neima Jahromi, The Fight for the Future of YouTube, NEW YORKER (July 8, 2019), https://www.newyorker.com/tech/annals-of- technology/the-fight-for-the-future-of-youtube. 114 GONZALEZ V. GOOGLE My view is consistent with our decision in Fields v. Twitter. In Fields, we acknowledged that acts of international terrorism are foreseeable consequences of financial support to a terrorist organization, but we also noted that such fungibility “does not relieve claimants of their burden to show causation.” Id. at 749. Fields requires that a plaintiff plausibly allege a “direct relationship between a defendant’s act and [a plaintiff’s] injur[ies],” id. at 748, and that element is met here because there is a sufficient nexus. Plaintiffs allege that ISIS operatives involved in the Paris Attacks posted links to ISIS YouTube videos. The sum of Plaintiffs’ allegations demonstrate that the terrorists responsible for Plaintiffs’ injuries used YouTube as an integral component of recruiting, and that such recruiting is necessary to carry out attacks at the scale of those in Paris. Specifically, Plaintiffs allege that at least two of the twelve ISIS terrorists who carried out the Paris Attacks, Abaaoud and Laachraoui, used online social media platforms to post links to ISIS recruitment YouTube videos and “jihadi YouTube videos.” Plaintiffs allege that Abaaoud, “considered the operational leader of the Paris Attack,” was an active user of social media, including YouTube. In a March 2014 ISIS YouTube video, “Abaaoud gave a monologue (in French) recruiting jihadi fighters for ISIS.” Plaintiffs also allege that at the time of the attacks these two ISIS terrorists, who were “instrumental in the Paris Attack,” were members of or at least involved with ISIS networks in Belgium called “The Zerkani Network” and Sharia4Belgium. The Belgian networks “used and relied on social media to build and maintain connections with ISIS recruits.” Plaintiffs allege that there was a pervasive network GONZALEZ V. GOOGLE 115 of ISIS recruiters in Belgium, which has been called “the epicenter of the Islamic State’s efforts to attack Europe.” Sharia4Belgium maintained several active YouTube channels, still active at the time of the Paris Attacks, “which it used to post sermons, speeches, news events, and other materials to lure, recruit, and indoctrinate young Muslims to travel to Syria and Iraq to join ISIS.” Plaintiffs allege that there was significant overlap and coordination over time between Sharia4Belgium and “The Zerkani Network.” Plaintiffs allege that Laachraoui was involved with Sharia4Belgium at the time of the Paris Attacks, and his social media accounts appear to show that he followed ISIS social media and posted links to jihadi YouTube videos on his own account. Though Plaintiffs do not specifically allege how the perpetrators of the Paris Attack were radicalized, such an allegation is not necessary to plausibly state their claim. It is enough that the complaint alleged that the perpetrators themselves actively used YouTube to recruit others to ISIS, gaining resources with which to plan and implement their attacks; absent the participation of the social media companies for their own profit-centered purposes, terrorist groups like ISIS would not have these resources. Additionally, Plaintiffs alleged that “The Zerkani Network” recruited one of the shooters, Abaaoud, “an active user of social media, including YouTube,” and also alleged that the network “used and relied on social media” to recruit, permitting the inference that it is probable Abbaoud was radicalized through social media. Viewing these allegations in the light most favorable to the nonmoving party, as we must, Campidoglio, 870 F.3d at 970, Plaintiffs have plausibly alleged a sufficient nexus between Google’s conduct and the 116 GONZALEZ V. GOOGLE Paris Attack victims’ injuries to satisfy a proximate cause threshold standard.^[11] A possible analogy may help to illustrate how the social media companies’ enhancement and spread of ISIS propaganda promoting violence and seeking to convert recruits has a direct relation to the damages caused here. Let’s assume that a person on one side of a crowded football stadium fires a high-powered rifle aimed at a crowd on the opposite side of the stadium, filled with people, though all identities are unclear. Would the majority here say that the rifle shot striking an unidentified viewer on the other side of the stadium had no “direct relation” to the shooter and that the shot did not proximately cause a resulting death? I think not. There is direct relation between shooter and victim there sufficient to satisfy Fields and there is similar direct relation here between the challenged conduct of the Defendant social media companies and the victims of ISIS violence in these cases to say that the challenged conduct, if shown to be illegal, was a proximate cause of damages. ^[11] It is worth noting that the contrary conclusion, espoused by the majority, would put these and future plaintiffs in an untenable position. If we required plaintiffs to specify exactly how an individual terrorist became radicalized without the benefit of discovery, then it is unlikely that any such claims could go forward. At the motion to dismiss stage, with notice pleading principles in mind, the Gonzalez Plaintiffs need only plausibly allege “some direct relation” between the terrorist’s actions and the social media companies’ conduct. See Fields, 881 F.3d at 749 (citation omitted). Here, Plaintiffs alleged that the perpetrator of the Paris Attack was a member of a particular network that used social media to recruit its members, and that the perpetrator himself was a regular user of social media. Given that it is unlikely potential terrorists will announce the avenues by which they were radicalized, such inferences are permissible. GONZALEZ V. GOOGLE 117 IV I next turn to whether Plaintiffs have adequately alleged claims against Google for secondary liability under JASTA. As with primary liability, the majority addressed only the revenue sharing claims in its opinion, but I would hold that for either set of claims, Plaintiffs have successfully stated a claim for secondary liability. Congress amended the ATA by enacting JASTA in 2016, Pub. L. No. 144-222, 130 Stat. 854 (Sept. 28, 2016), which extends liability to persons who aid and abet by providing substantial assistance to persons who commit acts of international terrorism, and those who conspire to commit such acts. ^[18] U.S.C. § 2333(d)(2). Under § 2333(d)(2) of the ATA, “liability may be asserted as to any person who aids and abets, by knowingly providing substantial assistance” to “the person who committed . . . an act of international terrorism.” Id. I recognize the proper legal framework for analyzing such claims as that described in Halberstam v. Welch, 705 F.2d 472 (D.C. Cir. 1983). Like the majority, I first conclude that the first two Halberstam factors have been satisfied here: (1) the party whom the defendant aids performed a wrongful act that caused an injury; and (2) the defendant was “generally aware of his role as part of an overall illegal or tortious activity at the time that he provide[d] the assistance.” Halberstam, 705 F.2d at 477. For the first element, the complaint plausibly alleges that the Paris Attacks were “committed, planned, or authorized” by ISIS, a designated terrorist organization. See 18 U.S.C. § 2333(d)(2). For the second element, I agree that Google was “generally aware of its role in ISIS’s terrorist activities” at the time it used its content-generating algorithms to send a message to YouTube users and at the time it shared revenue 118 GONZALEZ V. GOOGLE through AdSense. In both cases, Google was aware that it assumed a role in ISIS’s terrorist activities. See Halberstam, 705 F.2d at 488; see also Linde, 882 F.3d at 329 (noting that the element does not require a showing of “specific intent” as in criminal aiding and abetting, nor does it require that the defendant “knew of the specific attacks at issue”). Unlike my colleagues, however, I also conclude that the final element is met: the defendant “knowingly and substantially assisted[ed] the principal violation.” Halberstam, 705 F.2d at 488. The majority acknowledges that Google knowingly assisted the principal violation, but denies that such assistance was “substantial.” 12 I would hold that Google’s assistance via its content- generating algorithms and revenue sharing was both knowing and substantial. I need not view the non-revenue sharing claims and revenue sharing claims in isolation in this portion of my analysis. Because I conclude that both sets of Plaintiffs claims are not barred by Section 230, it is the sum of Google’s conduct that must be considered when assessing whether the assistance was substantial. The Halberstam court identified six factors relevant to assessing whether the substantial assistance component is satisfied: “(1) the nature of the act encouraged, (2) the amount of assistance given by 12 It may be that what is considered by one person to be “substantial assistance” is considered by another merely de minimis or inconsequential. But even if that is so, it would be a better procedure to leave that decision to fairly selected jurors with proper jury instructions explaining the “substantial assistance” element. But to me it is clear that ISIS could not exist and renew itself without constant recruitment of foot soldiers to carry out its violent missions, often at the cost of their own lives, so I regret that I cannot persuade my colleagues here to adopt a more permissive standard for substantial assistance. GONZALEZ V. GOOGLE 119 defendant, (3) defendant’s presence or absence at the time of the tort, (4) defendant’s relation to the principal, (5) defendant’s state of mind, and (6) the period of defendant’s assistance.” Linde, 882 F.3d at 329 (citing Halberstam, 705 F.2d at 483–84). Under the first factor, the Halberstam court emphasized that the nature of the principal’s act “dictates what aid might matter, i.e., be substantial.” 705 F.2d at 484. The remaining factors must be viewed through this lens. ISIS’s long-running and far-ranging terrorist campaign depends on the continued provision of money and recruits. Google provided both. As the majority acknowledges, financial support is “indisputably important” to operating a terrorism campaign, and any money provided to the organization may aid its goals. See id. at 488; Fields, 881 F.3d at 748. The majority also acknowledges, in the context of reversing the district court’s dismissal of Taamneh, that YouTube videos encourage ISIS’s terrorism campaign—an enterprise that is “heavily dependent on social media platforms to recruit members, to raise funds, and to disseminate propaganda.” Google provided free exposure to a dangerous organization, thereby facilitating ISIS’s ability to reach and rouse prospective recruits. The Gonzalez complaint alleges that ISIS through YouTube exaggerated its territorial expansion by disseminating videos with maps showing ISIS’s claims that it controlled certain regions where other groups had pledged allegiance to ISIS. The fourth factor also weighs in favor of recognizing substantial assistance: defendant’s “relation” to the principal—or the extent to which an entity “may possess greater powers of suggestion.” YouTube’s role in cultivating extremist behavior has been widely acknowledged and the platform reaches a virtually unlimited number of potential recruits due to the ubiquity of the Internet. The sixth factor, “duration of 120 GONZALEZ V. GOOGLE the assistance provided,” concerns the length of time an alleged aider and abettor has been involved with the tortfeasor. See Halberstam, 705 F.2d at 484 (emphasis omitted). Though the complaint in Gonzalez lacks specific evidence about the length of time Google provided assistance to ISIS, Plaintiffs allege the placement of ISIS recruiting videos going back at least four years before the Paris Attacks, in 2014. The complaint also alleged through news sources that advertisements were placed on ISIS’s YouTube videos as early as March 2015, three years before the Paris Attacks. I would hold that years of hosting ISIS content and providing it with a percentage of revenue is sufficient duration. Though Plaintiffs do not allege that Google shared ISIS’s terrorist goals, Halberstam also directs that under the fifth factor, defendant’s “state of mind,” the court can consider the duration factor because it “almost certainly affects the quality and extent” of the aid, the amount of aid provided, and “it may afford evidence of the defendant’s state of mind.” Id. Even considering state of mind on its own and viewing that factor in light of “the nature of the act encouraged,” see id., providing financial assistance and exposure to—to put it mildly—a dangerous group, is sufficient for state of mind to weigh against Google. As I see it, the conduct of Google, Twitter, and Facebook as related to the risks of terrorist attacks by ISIS, absent their more active review and policing of sites, is either recklessly indifferent or willfully blind, as they enjoy increased advertising revenue associated with eyeballs on videos or posts about ISIS attacks. Taken as true and viewed in the light most favorable to Plaintiffs, I would hold that these allegations establish that Google’s assistance was sufficiently “substantial” for purposes of § 2333(d)(2). These same considerations apply GONZALEZ V. GOOGLE 121 in all three cases, so in each I would hold there was substantial assistance for purposes of § 2333(d)(2). I add a brief comment about the Clayborn v. Twitter case. There the majority would uphold dismissal of the claims because of its view that Plaintiffs do not plausibly allege that ISIS “committed, planned, or authorized” the San Bernardino attack, as is required under 18 U.S.C. § 2333(d)(2). The majority relies on a Sixth Circuit decision, Crosby v Twitter, Inc., 921 F.3d 617 (6th Cir. 2019), but its reasoning is not persuasive and does not bind or even guide our circuit, because there, the complaint produced “no allegations that ISIS was involved with the Pulse Night Club shooting.” Id. at 626. However, the record here is distinctly and plainly to the contrary: The complaint expressly alleges that prior to or during the attack, one of the perpetrators—Tashfeen Malik—declared on her Facebook page the two shooters’ allegiance and loyalty to an ISIS leader. Two days after the attack, ISIS issued a statement on a radio station claiming responsibility for the attack. The FBI confirmed that one of the shooters, a few years before the attack, had face-to-face meetings with five people known to have “links to terrorism.” Further, Plaintiffs allege that FBI investigators found an explosive device placed at the crime scene that was likely intended to be detonated by the arrival of first responders. A Department of Justice report described this as “a frequent, well documented practice in international terrorism incidents.” Importantly, FBI investigators explained that this “terrorist tactic ha[d] been outlined in Al Qaeda’s Inspire Magazine, as well as in ISIS’s Dabiq Magazine.” Plaintiffs allege that these magazines are disseminated on Defendants’ platforms. Together, these allegations permit the fair inference that the attack which was planned for at least one year was inspired by—and implicitly authorized by—ISIS. 122 GONZALEZ V. GOOGLE In my view, even if Malik had been “self-radicalized” without direct communications or meetings with ISIS operatives, Plaintiffs plausibly allege that the self- radicalization process included exposure to the violent recruiting videos of ISIS, along with lectures from incendiary advocates of violence against non-believers. According to the complaint, in Senate Judiciary Committee testimony, then-FBI Director James Comey described the pair as having “consum[ed] poison on the internet” and been “radicalized to jihadism and to martyrdom via social media platforms available to them.” Finally, even assuming the perpetrators had little advance connection with ISIS, well-established principles of agency law illustrate that authorization can occur not only by advance planning, but also by ratification. See Restatement (Third) of Agency, § 4.01(1) (1933) (defining ratification as “the affirmance of a prior act done by another, whereby the act is given effect as if done by an agent acting with actual authority”).13 Because the San Bernardino 13 Contrary to the majority’s contention, there is support for applying common law agency principles to secondary liability for acts of international terrorism. For one thing, “statutes are presumed not to disturb the common law, ‘unless the language of a statute [is] clear and explicit for this purpose.’” State Eng’r of Nev. v. S. Fork Band of Te- Moak Tribe of W. Shoshone Indians of Nev., 339 F.3d 804, 814 (9th Cir. 2003) (quoting Norfolk Redevelopment & Hous. Auth. v. Chesapeake & Potomac Tel. Co. of Va., 464 U.S. 30, 35 (1983)). In my view, nothing in the statute precludes consideration of common law principles. Second, the Supreme Court has stated that apparent authority principles “ha[ve] long been the settled rule in the federal system.” Am. Soc’y of Mechanical Eng’rs, Inc. v. Hydrolevel Corp., 456 U.S. 556, 567 (1982). Section 2333(d)(2) assigns liability for injuries arising from acts of international terrorism, where that act was authorized by a terrorist organization. See 18 U.S.C. § 2333(d)(2). In my view, asking whether a terrorist organization authorized a particular terrorist act is properly viewed under GONZALEZ V. GOOGLE 123 shooters pledged themselves to ISIS before or during the attack, and an act is ratifiable “if the actor acted or purported to act as an agent on the person’s behalf,” id. § 4.03, the attack can be considered authorized by ISIS. For the foregoing reasons, the complaint in Clayborn makes allegations sufficient to state a claim for liability under the ATA. V In my view, the claims asserted in the three complaints on appeal should all be sustained and permitted to go forward in discovery based on the statutory law standards above discussed. But even if I am incorrect in my view of the governing statutory law, those claims should be able to go forward with complaint amendment based on a still extant specialized federal common law in aid of national security against terrorism. After the general common law regime of Swift v. Tyson was overruled by Erie, a sphere of specialized federal common law remains and could support Plaintiffs’ claims here. See e.g., 19 Charles Alan Wright & Arthur R. Miller, Federal Practice & Procedure § 4514 (3d ed. 2021). As the Wright & Miller treatise explains, “the federal common law that has developed since Erie differs from the federal general common law [rejected in] Swift v. Tyson because it falls within an area of federal or national competence.” Id. (footnote omitted); see also 17A Moore’s Federal Practice, Civil § 124.40 (2020). Many federal court precedents have applied these principles, which are particularly well-suited when claims involve an area of the common-law agency framework as a question of whether the perpetrator was acting as an agent of the terrorist organization. 124 GONZALEZ V. GOOGLE heightened federal interest, such as international terrorism, or when gaps exist in a federal regulatory scheme. E.g., Boyle v. United Techs. Corp., 487 U.S. 500, 507–08 (1988); Textile Workers Union of Am. v. Lincoln Mills of Ala., 353 U.S. 448, 456–57 (1957); King Jewelry, Inc. v. Fed. Express Corp., 316 F.3d 961, 964–65 (9th Cir. 2003).14 Also, our court should not ignore other potential areas of human conduct that can be negatively impacted by an unregulated social media regime, coupled with efforts by groups hostile to the idea of American democracy to use social media in order to divide or terrorize our public. Areas of particular concern include impacts of social media in realms such as election law, the laws governing public order and protest, and even insurrection. We should not of course ignore the tremendous, indeed almost unquantifiable, benefits to the public from social media. Social media permits friends to stay in contact, as for example with a club or group from high school or college, lets people make new friends, or even lets people see or be exposed to new sights from different parts of the world. People met through social media, who may have different interests, perspectives, and priorities from other social media users, can in many cases enrich those users’ lives. Places visited on the internet, often encouraged or directed through social media, can serve the same benign function. But at the 14 Contrary to the majority’s contention, my view on when federal common law may be created is narrow. Though the federal courts “are not free to manufacture entirely new causes of action merely because the political branches have not acted,” I believe that we can act where gaps are present in an existing federal statutory scheme and the claim involved is one of unique federal concern. GONZALEZ V. GOOGLE 125 same time, benefit alone cannot end the inquiry. Social media activities also carry with them some risks and detriments to the public. For example, there is no doubt that modern pharmaceutical drugs give benefits to the public that were impossible at earlier times and are greatly valued by those who use them. But drugs can also have harmful impacts and, accordingly, they are regulated by the Food & Drug Administration. Similarly, modern aircraft help people move from one part of our world to another with great speed and ease, but we regulate airlines through the Federal Aviation Administration. One could go on and on as almost every major activity in the modern world faces some type of federal regulation. This regulation of the social media companies would best be examined by congressional committees with subpoena power and the ability to create new regulatory laws if needed and desirable. Or the government could create a new federal agency or Board or add powers or some supplemental standards to an existing federal agency, leaving the regulation of social media in part to a federal executive agency that is committed to bringing its technical expertise and knowledge of any areas of specialized federal concerns such as international terrorism and threats to democracy to bear on this issue. A specialized federal agency could call witnesses for testimony, assist meaningfully in a congressional task to prepare appropriate legislative guidance or prohibitions, have investigators to look into areas of concern, establish regulatory standards, and possibly also include an arm to enforce the law and its standards. See, e.g., Myers v. United States, 272 U.S. 52, 129 (1926) (recognizing congressional authority to create federal agencies and define their scope and jurisdiction). 126 GONZALEZ V. GOOGLE VI These cases, and others like them pending in the federal courts, try for basic justice, but there is a fundamental question whether the federal courts are best suited to deliver it. I conclude with the following thoughts. First, it would be preferable if the political branches of government, the legislature or the Executive Branch, would seriously grapple with the issue of unregulated social media power being used to amplify or to distort views asserted by users, and sometimes even by hostile nations using social media to wage asymmetric warfare or to impair democracy. But if Congress continues to sleep at the switch of social media regulation in the face of courts broadening what appears to have been its initial and literal language and expressed intention under Section 230, then it must fall to the federal courts to consider rectifying those errors itself by providing remedies to those who are injured by dangerous and unreasonable conduct. Second, it would be preferable if the social media companies monitored their own activities sufficiently to protect the public, but in my view, to date they have not done that. It was one thing, at the dawn of the Internet era, to give protection to Internet companies to facilitate growth. But it is quite another thing to provide broad immunity at a time such as now when such companies are remarkably large and with massive staffs and perhaps the best technical abilities. It is not realistic to anticipate that social media companies will self-police adequately in the face of their incentives to maximize profits by maximizing advertising revenues, which means increasing the eyeballs directed to their websites. The large corporations controlling the platforms at issue in these GONZALEZ V. GOOGLE 127 appeals can instead be expected to act in their own best financial interest, and to me, it makes absolutely no sense to leave such decisions to the self-interested proclamations of CEOs or other employees of the various social media companies.^[15] Society for centuries has known that it is folly to ask the fox to guard the henhouse. Third, the problem with a lack of social media regulation goes even beyond the dreadfully important subject of terrorism. Indeed, in connection with 21st-century political elections, some commenters have expressed concerns that social media has the ability to distort and tribalize public opinion, to spread falsehoods as well as truth, and to funnel like-minded news reports to groups in a way that makes them think there are “alternative facts” or “competing realities” that exist, rather than recognize more correctly that there are “truth” and “lies.”16 Fourth, to the extent any of our Ninth Circuit precedent stands in the way of a sensible resolution of claims like those presented on appeal here, where terrorist organizations like ISIS have obviously played Google and YouTube like a fiddle, then in my view we should take these or other related cases en banc to give a full review. ^[15] E.g., 1 ADAM SMITH, AN INQUIRY INTO THE NATURE AND CAUSES OF THE WEALTH OF NATIONS 13 (1776) (“It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest.”). 16 See Ross Douthat, Why Do So Many Americans Think the Election Was Stolen?, N.Y. TIMES (Dec. ^[5], 2020), https://www.nytimes.com/2020/12/05/opinion/sunday/trump-election- fraud.html?smid=tw-share. 128 GONZALEZ V. GOOGLE Fifth, because the issues are difficult and only the Supreme Court can speak with authority ultimately on federal law, it would be desirable for the Supreme Court to take up the subject of Section 230 immunity and perhaps any related First Amendment issues, to the extent claims relating to terrorist speech are properly considered under that framework. Justice Oliver Wendell Holmes, Jr. made famous and enshrined in our law the idea that: “The life of the law has not been logic, it has been experience.” OLIVER WENDELL HOLMES, JR., THE COMMON LAW, Lecture I (1881). But when almost all claims against social media companies are dismissed at the outset because of an overbroad view of Section 230 immunity, how is society to develop the experience that can guide its development of law in a sensible way that protects people from undue harm? Justice Holmes also developed the idea that speech should not be constrained absent “clear and present danger,” see Schenck v. United States, 249 U.S. 47 (1919). To some degree this test still resounds in our First Amendment law. See United States v. Alvarez, 617 F.3d 1198, 1214 (9th Cir. 2010). A variation on this view culminated in Brandenburg v. Ohio, 395 U.S. 444 (1969), where the Supreme Court suggested that imminent lawless action was necessary before speech should be constrained. But perhaps given the current state of society, and the catastrophic dangers to the public that can be posed by terrorist activities, public safety may require that speech be limited when it poses a clear and increasing or gathering danger, rather than only “imminent” danger as reflected in Brandenburg, which I consider the Supreme Court’s last word on this subject. GONZALEZ V. GOOGLE 129 I also note that Oliver Wendell Holmes, Jr.’s famous pen pal and intellectual collaborator, Sir Frederick Pollock,17 in his beginning primer of the law of torts, suggested that a principal force underlying all the varied types of tort cases was the desire of courts to provide a doctrinal basis for remedy in the case of injuries from harmful and unreasonable conduct. Pollock suggested that a “tort is an act or omission (not merely the breach of a duty arising out of a personal relation, or undertaken by contract) which is related to harm suffered by a determinate person in one of the following ways.” See SIR FREDERICK POLLOCK, THE LAW OF TORTS: A TREATISE ON THE PRINCIPLES OF OBLIGATIONS ARISING FROM CIVIL WRONGS IN THE COMMON LAW 20 (4th ed. 1895). Among those ways a person can be harmed were these two, which are pertinent in assessing whether Plaintiffs’ claims can be asserted as part of a federal common law: “(c) it may be an act or omission causing harm which the person so acting or so omitting did not intend to cause, but might and should with due diligence have foreseen and prevented,” and “(d) it may in special cases consist in not avoiding or preventing that which the party was bound, absolutely or within limits to avoid or prevent.” Id. Here, it could be expected that through federal common law development or statutory positive law, the social media companies will be held to some reasonable standard of conduct when they have 17 See Oliver Wendell Holmes Jr. & Sir Frederick Pollock, Holmes–Pollock Letters: The Correspondence of Mr Justice Holmes and Sir Frederick Pollock, 1874–1932 (2d ed. 1961). 130 GONZALEZ V. GOOGLE failed to regulate their own actions in the interests of the public.^[18] As a matter of federal common law, I would hold that when social media companies in their platforms use systems or procedures that are unreasonably dangerous to the public—as in the case where their systems line up repeated messages in aid of terrorists like ISIS—or when they omit to act to avoid harm when omitting the act is unreasonably dangerous to the public—as in the case where they fail to review and self-regulate their websites adequately to notice and remove propaganda videos from ISIS that are likely to cause harm—then there should be a federal common law claim available against them. Consider the most widely used standard for products liability cases. See Restatement (Second) of Torts, § 402A (1965). This suggests that manufacturers are responsible in tort if they make unreasonably dangerous products that cause individual or social harm. Section 402A states: “One who sells any product in a defective condition unreasonably dangerous to the user or consumer or to his property is subject to liability for physical harm thereby caused” to the user or a third party. Id. Here and similarly, social media companies should be viewed as making and “selling” their social media products through the device of forced advertising under the eyes of users. Viewed in this light, they should be tested under a federal tort principle with a standard similar to and adapted from this Restatement language under a federal common law 18 Developing federal common law on these issues will require the diligent and combined efforts of the federal courts and of legal scholars. See, e.g., Hon. Wade H. McCree, Jr., The Annual John Randolph Tucker Lecture, Partners in a Process: The Academy and the Courts, 37 WASH. & LEE. L. REV. 1041 (1981). GONZALEZ V. GOOGLE 131 development. If social media companies use “neutral” algorithms that cause unreasonably dangerous consequences, under proper standards of law with limiting jury instructions, they might be held responsible. Developing a federal common law standard would be superior to merely dismissing all claims against social media companies based on an over- broad interpretation of Section 230 delivering a blanket immunity, which in my view is inconsistent with congressional intent and detrimental to the interests of the general public. 132 GONZALEZ V. GOOGLE ATTACHMENT A ȱ ȱ KATZMANN,ȱChiefȱJudge,ȱconcurringȱinȱpartȱandȱdissentingȱinȱpart:ȱȱ Iȱagreeȱwithȱmuchȱofȱtheȱreasoningȱinȱtheȱexcellentȱmajorityȱopinion,ȱandȱIȱ joinȱthatȱopinionȱexceptȱforȱPartsȱIȱandȱIIȱofȱtheȱDiscussion.ȱButȱIȱmustȱrespectfullyȱ partȱcompanyȱwithȱtheȱmajorityȱonȱitsȱtreatmentȱofȱFacebook’sȱfriendȬȱandȱcontentȬ suggestionȱalgorithmsȱunderȱtheȱCommunicationsȱDecencyȱActȱ(“CDA”).1ȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 1ȱIȱagreeȱwithȱtheȱmajorityȱthatȱtheȱCDA’sȱexceptionȱforȱenforcementȱofȱcriminalȱ laws,ȱ47ȱU.S.C.ȱ§ȱ230(e)(1),ȱdoesȱnotȱapplyȱtoȱplaintiffs’ȱclaims,ȱseeȱante,ȱatȱ50Ȭ54.ȱHowever,ȱ Iȱfindȱtheȱquestionȱtoȱbeȱsomewhatȱcloserȱthanȱtheȱmajorityȱdoes,ȱinȱpartȱbecauseȱsomeȱofȱ theȱstatutesȱenumeratedȱinȱ§ȱ230(e)(1)ȱthemselvesȱcontainȱcivilȱremedies.ȱSectionȱ230(e)(1)ȱ statesȱthatȱ“[n]othingȱinȱ[§ȱ230]ȱshallȱbeȱconstruedȱtoȱimpairȱtheȱenforcementȱofȱsectionȱ 223ȱ orȱ 231ȱ ofȱ thisȱ title,ȱ chapterȱ 71ȱ (relatingȱ toȱ obscenity)ȱ orȱ 110ȱ (relatingȱ toȱ sexualȱ exploitationȱofȱchildren)ȱofȱtitleȱ18,ȱorȱanyȱotherȱFederalȱcriminalȱstatute.”ȱOneȱofȱthoseȱ enumeratedȱchapters—Chapterȱ110ȱofȱTitleȱ18—includesȱaȱcivilȱsuitȱprovisionȱforȱvictimsȱ ofȱ specificȱ childȱ sexȱ crimes.ȱ Seeȱ 18ȱ U.S.C.ȱ §ȱ 2255.ȱ Meanwhile,ȱ 47ȱ U.S.C.ȱ §ȱ 223—whichȱ prohibitsȱ obsceneȱ orȱ harassingȱ phoneȱ calls—specifiesȱ thatȱ civilȱ finesȱ mayȱ beȱ leviedȱ “pursuantȱtoȱcivilȱactionȱby,”ȱorȱ“afterȱappropriateȱadministrativeȱproceedings”ȱof,ȱtheȱ FederalȱCommunicationsȱCommissionȱ(“FCC”),ȱandȱitȱauthorizesȱtheȱAttorneyȱGeneralȱ toȱ bringȱ civilȱ suitsȱ toȱ enjoinȱ practicesȱ thatȱ violateȱ theȱ statute.ȱ 47ȱ U.S.C.ȱ §ȱ223(b)(5)(B)Ȭ (b)(6).ȱ Ifȱ §ȱ 230(e)(1)ȱ coversȱ “enforcement”ȱ ofȱ theȱ listedȱ chaptersȱ inȱ theirȱ entirety,ȱ itȱ isȱ difficultȱ toȱ seeȱ howȱ itȱ wouldȱ notȱ coverȱ otherȱ provisionsȱ thatȱ authorizeȱ civilȱ suitsȱ forȱ violationsȱofȱcriminalȱlaws,ȱparticularlyȱgivenȱthatȱtheȱenumeratedȱlistȱisȱfollowedȱbyȱ“orȱ anyȱotherȱcriminalȱlaw.”ȱȱ ȱ However,ȱ asȱ detailedȱ post,ȱ §ȱ 230ȱ wasȱ designedȱ asȱ aȱ privateȬsectorȬdrivenȱ alternativeȱ toȱ aȱ Senateȱ planȱ thatȱ wouldȱ allowȱ theȱ FCCȱ “eitherȱ civillyȱ orȱ criminally,ȱ toȱ punishȱ people”ȱ whoȱ putȱ objectionableȱ materialȱ onȱ theȱ Internet.ȱ 141ȱ Cong.ȱ Rec.ȱ 22,045ȱ (1995)ȱ(statementȱofȱRep.ȱCox);ȱaccordȱid.ȱatȱ22,045Ȭ46ȱ(statementȱofȱRep.ȱWyden);ȱseeȱRenoȱ v.ȱACLU,ȱ521ȱU.S.ȱ844,ȱ859ȱ&ȱn.24ȱ(1997).ȱOnȱtheȱHouseȱfloor,ȱauthorȱChristopherȱCoxȱ disparagedȱ theȱ ideaȱ ofȱ FCCȱ enforcementȱ andȱ thenȱ stated:ȱ “Certainly,ȱ criminalȱ ȱ enforcementȱofȱourȱobscenityȱlawsȱasȱanȱadjunctȱisȱaȱusefulȱwayȱofȱpunishingȱtheȱtrulyȱ 1ȱ ȱ Asȱ toȱ theȱ reasonsȱ forȱ myȱ disagreement,ȱ considerȱ aȱ hypothetical.ȱ Supposeȱ thatȱyouȱareȱaȱpublishedȱauthor.ȱOneȱday,ȱanȱacquaintanceȱcalls.ȱ“I’veȱbeenȱreadingȱ overȱeverythingȱyou’veȱeverȱpublished,”ȱheȱinformsȱyou.ȱ“I’veȱalsoȱbeenȱlookingȱ atȱeverythingȱyou’veȱeverȱsaidȱonȱtheȱInternet.ȱI’veȱdoneȱtheȱsameȱforȱthisȱotherȱ author.ȱ Youȱ twoȱ haveȱ veryȱ similarȱ interests;ȱ Iȱ thinkȱ you’dȱ getȱ along.”ȱ Theȱ acquaintanceȱ thenȱ givesȱ youȱ theȱ otherȱ author’sȱ contactȱ informationȱ andȱ photo,ȱ alongȱwithȱaȱlinkȱtoȱallȱherȱpublishedȱworks.ȱHeȱcallsȱbackȱthreeȱmoreȱtimesȱoverȱ theȱnextȱweekȱwithȱmoreȱnamesȱofȱwritersȱyouȱshouldȱgetȱtoȱknow.ȱȱ Now,ȱyouȱmightȱsayȱyourȱacquaintanceȱfanciesȱhimselfȱaȱmatchmaker.ȱButȱ wouldȱyouȱsayȱhe’sȱactingȱasȱtheȱpublisherȱofȱtheȱotherȱauthors’ȱwork?ȱ Facebookȱ andȱ theȱ majorityȱ wouldȱ haveȱ usȱ answerȱ thisȱ questionȱ “yes.”ȱ I,ȱ however,ȱcannotȱdoȱso.ȱForȱtheȱscenarioȱIȱhaveȱjustȱdescribedȱisȱlittleȱdifferentȱfromȱ howȱFacebook’sȱalgorithmsȱallegedlyȱwork.ȱAndȱwhileȱthoseȱalgorithmsȱdoȱendȱ upȱshowingȱusersȱprofile,ȱgroup,ȱorȱeventȱpagesȱwrittenȱbyȱotherȱusers,ȱitȱstrainsȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ guilty.”ȱ141ȱCong.ȱRec.ȱ22,045ȱ(emphasisȱadded).ȱThisȱhistory,ȱalongȱwithȱtheȱprovision’sȱ title,ȱstronglyȱsuggestsȱthatȱ§ȱ230(e)(1)ȱwasȱintendedȱasȱaȱnarrowȱcriminalȬlawȱexception.ȱ Itȱwouldȱbeȱodd,ȱthen,ȱtoȱreadȱ §ȱ230(e)(1)ȱasȱallowingȱforȱcivilȱenforcementȱby,ȱamongȱ ȱothers,ȱtheȱFCC,ȱevenȱifȱonlyȱinȱaidȱofȱcriminalȱlawȱenforcement.ȱ 2ȱ ȱ theȱEnglishȱlanguageȱtoȱsayȱthatȱinȱtargetingȱandȱrecommendingȱtheseȱwritingsȱtoȱ users—andȱ therebyȱ forgingȱ connections,ȱ developingȱ newȱ socialȱ networks— Facebookȱ isȱ actingȱ asȱ “theȱ publisherȱ ofȱ .ȱ .ȱ .ȱ informationȱ providedȱ byȱ anotherȱ informationȱcontentȱprovider.”ȱ47ȱU.S.C.ȱ§ȱ230(c)(1)ȱ(emphasisȱadded).ȱȱ Itȱ wouldȱ beȱ oneȱ thingȱ ifȱ congressionalȱ intentȱ compelledȱ usȱ toȱ adoptȱ theȱ majority’sȱ reading.ȱ Itȱ doesȱ not.ȱ Instead,ȱ weȱ todayȱ extendȱ aȱ provisionȱ thatȱ wasȱ designedȱtoȱencourageȱcomputerȱserviceȱprovidersȱtoȱshieldȱminorsȱfromȱobsceneȱ materialȱsoȱthatȱitȱnowȱimmunizesȱthoseȱsameȱprovidersȱforȱallegedlyȱconnectingȱ terroristsȱtoȱoneȱanother.ȱNeitherȱtheȱimpetusȱforȱnorȱtheȱtextȱofȱ§ȱ230(c)(1)ȱrequiresȱ suchȱaȱresult.ȱWhenȱaȱplaintiffȱbringsȱaȱclaimȱthatȱisȱbasedȱnotȱonȱtheȱcontentȱofȱtheȱ informationȱ shownȱ butȱ ratherȱ onȱ theȱ connectionsȱ Facebook’sȱ algorithmsȱ makeȱ betweenȱindividuals,ȱtheȱCDAȱdoesȱnotȱandȱshouldȱnotȱbarȱrelief.ȱȱ TheȱAntiȬTerrorismȱActȱ(“ATA”)ȱclaimsȱinȱthisȱcaseȱfitȱthisȱbill.ȱAccordingȱ toȱplaintiffs’ȱProposedȱSecondȱAmendedȱComplaintȱ(“PSAC”)—whichȱweȱmustȱ takeȱ asȱ trueȱ atȱ thisȱ earlyȱ stage—Facebookȱ hasȱ developedȱ “sophisticatedȱ algorithm[s]”ȱ forȱ bringingȱ itsȱ usersȱ together.ȱ App’xȱ 347ȱ ¶ȱ 622.ȱ Afterȱ collectingȱ mountainsȱ ofȱ dataȱ aboutȱ eachȱ user’sȱ activityȱ onȱ andȱ offȱ itsȱ platform,ȱ Facebookȱ 3ȱ ȱ ȱ unleashesȱitsȱalgorithmsȱtoȱgenerateȱfriend,ȱgroup,ȱandȱeventȱsuggestionsȱbasedȱ onȱwhatȱitȱperceivesȱtoȱbeȱtheȱuser’sȱinterests.ȱId.ȱatȱ345Ȭ46ȱ¶¶ȱ608Ȭ14.ȱIfȱaȱuserȱpostsȱ aboutȱaȱHamasȱattackȱorȱsearchesȱforȱinformationȱaboutȱaȱHamasȱleader,ȱFacebookȱ mayȱ“suggest”ȱthatȱthatȱuserȱbecomeȱfriendsȱwithȱHamasȱterroristsȱonȱFacebookȱ orȱ joinȱ HamasȬrelatedȱ Facebookȱ groups.ȱ Byȱ “facilitat[ing]ȱ [Hamas’s]ȱ abilityȱ toȱ reachȱ andȱ engageȱ anȱ audienceȱ itȱ couldȱ notȱ otherwiseȱ reachȱ asȱ effectively,”ȱ plaintiffsȱ allegeȱ thatȱ Facebook’sȱ algorithmsȱ provideȱ materialȱ supportȱ andȱ personnelȱtoȱterrorists.ȱId.ȱatȱ347ȱ¶ȱ622;ȱseeȱid.ȱatȱ352Ȭ58ȱ¶¶ȱ646Ȭ77.ȱAsȱappliedȱtoȱtheȱ algorithms,ȱplaintiffs’ȱclaimsȱdoȱnotȱseekȱtoȱpunishȱFacebookȱforȱtheȱcontentȱothersȱ post,ȱ forȱ decidingȱ whetherȱ toȱ publishȱ thirdȱ parties’ȱ content,ȱ orȱ forȱ editingȱ (orȱ failingȱtoȱedit)ȱothers’ȱcontentȱbeforeȱpublishingȱit.ȱInȱshort,ȱtheyȱdoȱnotȱrelyȱonȱ treatingȱFacebookȱasȱ“theȱpublisher”ȱofȱothers’ȱinformation.ȱInstead,ȱtheyȱwouldȱ holdȱFacebookȱliableȱforȱitsȱaffirmativeȱroleȱinȱbringingȱterroristsȱtogether.ȱ WhenȱitȱcomesȱtoȱFacebook’sȱalgorithms,ȱthen,ȱplaintiffs’ȱcausesȱofȱactionȱdoȱ notȱrunȱafoulȱofȱtheȱCDA.ȱBecauseȱtheȱcourtȱbelowȱdidȱnotȱpassȱonȱtheȱmeritsȱofȱ theȱATAȱclaimsȱpressedȱbelow,ȱIȱwouldȱsendȱthisȱcaseȱbackȱtoȱtheȱdistrictȱcourtȱtoȱ decideȱtheȱmeritsȱinȱtheȱfirstȱinstance.ȱTheȱmajority,ȱhowever,ȱcutsȱoffȱallȱpossibilityȱ 4ȱ ȱ ȱ forȱ reliefȱ basedȱ onȱ algorithmsȱ likeȱ Facebook’s,ȱ evenȱ ifȱ theseȱ orȱ futureȱ plaintiffsȱ couldȱproveȱaȱsufficientȱnexusȱbetweenȱthoseȱalgorithmsȱandȱtheirȱinjuries.ȱInȱlightȱ ofȱ today’sȱ decisionȱ andȱ otherȱ judicialȱ interpretationsȱ ofȱ theȱ statuteȱ thatȱ haveȱ generallyȱimmunizedȱsocialȱmediaȱcompanies—andȱespeciallyȱinȱlightȱofȱtheȱnewȱ realityȱthatȱhasȱevolvedȱsinceȱtheȱCDA’sȱpassage—Congressȱmayȱwishȱtoȱrevisitȱ theȱ CDAȱ toȱ betterȱ calibrateȱ theȱ circumstancesȱ whereȱ suchȱ immunizationȱ isȱ appropriateȱandȱinappropriateȱinȱlightȱofȱcongressionalȱpurposes.ȱ I. ȱ ToȱseeȱhowȱfarȱweȱhaveȱstrayedȱfromȱtheȱpathȱonȱwhichȱCongressȱsetȱusȱout,ȱ weȱmustȱconsiderȱwhereȱthatȱpathȱbegan.ȱWhatȱisȱnowȱ47ȱU.S.C.ȱ§ȱ230ȱwasȱaddedȱ asȱanȱamendmentȱtoȱtheȱTelecommunicationsȱActȱofȱ1996,ȱaȱstatuteȱdesignedȱtoȱ deregulateȱandȱencourageȱinnovationȱinȱtheȱtelecommunicationsȱindustry.ȱPub.ȱL.ȱ 104Ȭ104,ȱ§ȱ509,ȱ110ȱStat.ȱ56,ȱ56,ȱ137Ȭ39;ȱseeȱReno,ȱ521ȱU.S.ȱatȱ857.ȱCongressȱdevotedȱ muchȱ committeeȱ attentionȱ toȱ traditionalȱ telephoneȱ andȱ broadcastȱ media;ȱ byȱ contrast,ȱ theȱ Internetȱ wasȱ anȱ afterthought,ȱ addressedȱ onlyȱ throughȱ floorȱ amendmentsȱorȱinȱconference.ȱReno,ȱ521ȱU.S.ȱatȱ857Ȭ58.ȱOfȱtheȱmyriadȱissuesȱtheȱ emergingȱInternetȱimplicated,ȱCongressȱtackledȱonlyȱone:ȱtheȱeaseȱwithȱwhichȱtheȱ 5ȱ ȱ ȱ Internetȱ deliversȱ indecentȱ orȱ offensiveȱ material,ȱ especiallyȱ toȱ minors.ȱ Seeȱ TelecommunicationsȱActȱofȱ1996,ȱtit.ȱV,ȱsubtit.ȱA,ȱ110ȱStat.ȱatȱ133Ȭ39.ȱAndȱ§ȱ230ȱ providedȱoneȱofȱtwoȱalternativeȱwaysȱofȱhandlingȱthisȱproblem.ȱ TheȱactionȱbeganȱinȱtheȱSenate.ȱSenatorȱJamesȱJ.ȱExonȱintroducedȱtheȱCDAȱ onȱFebruaryȱ1,ȱ1995.ȱSeeȱ141ȱCong.ȱRec.ȱ3,203.ȱHeȱpresentedȱaȱrevisedȱbillȱonȱJuneȱ 9,ȱ1995,ȱ“[t]heȱheartȱandȱtheȱsoul”ȱofȱwhichȱwasȱ“itsȱprotectionȱforȱfamiliesȱandȱ children.”ȱId.ȱatȱ15,503ȱ(statementȱofȱSen.ȱExon).ȱTheȱExonȱAmendmentȱsoughtȱtoȱ reduceȱ theȱ proliferationȱ ofȱ pornographyȱ andȱ otherȱ obsceneȱ materialȱ onlineȱ byȱ subjectingȱ toȱ civilȱ andȱ criminalȱ penaltiesȱ thoseȱ whoȱ useȱ interactiveȱ computerȱ servicesȱtoȱmake,ȱsolicit,ȱorȱtransmitȱoffensiveȱmaterial.ȱId.ȱatȱ15,505.ȱ ȱ TheȱHouseȱofȱRepresentativesȱhadȱtheȱsameȱgoal—toȱprotectȱchildrenȱfromȱ inappropriateȱ onlineȱ material—butȱ aȱ veryȱ differentȱ senseȱ ofȱ howȱ toȱ achieveȱ it.ȱ Congressmenȱ Christopherȱ Coxȱ (RȬCalifornia)ȱ andȱ Ronȱ Wydenȱ (DȬOregon)ȱ introducedȱ anȱ amendmentȱ toȱ theȱ Telecommunicationsȱ Act,ȱ entitledȱ “Onlineȱ FamilyȱEmpowerment,”ȱaboutȱtwoȱmonthsȱafterȱtheȱrevisedȱCDAȱappearedȱinȱtheȱ Senate.ȱ Seeȱ id.ȱ atȱ 22,044.ȱ Makingȱ theȱ argumentȱ forȱ theirȱ amendmentȱ duringȱ theȱ Houseȱfloorȱdebate,ȱCongressmanȱCoxȱstated:ȱ 6ȱ ȱ ȱ Weȱ wantȱ toȱ makeȱ sureȱ thatȱ everyoneȱ inȱ Americaȱ hasȱ anȱ openȱ invitationȱandȱfeelsȱwelcomeȱtoȱparticipateȱinȱtheȱInternet.ȱButȱasȱyouȱ know,ȱthereȱisȱsomeȱreasonȱforȱpeopleȱtoȱbeȱwaryȱbecause,ȱasȱaȱTimeȱ Magazineȱcoverȱstoryȱrecentlyȱhighlighted,ȱthereȱisȱinȱthisȱvastȱworldȱ ofȱ computerȱ information,ȱ aȱ literalȱ computerȱ library,ȱ someȱ offensiveȱ material,ȱsomeȱthingsȱinȱtheȱbookstore,ȱifȱyouȱwill,ȱthatȱourȱchildrenȱ oughtȱnotȱtoȱsee.ȱ ȱ Asȱtheȱparentȱofȱtwo,ȱIȱwantȱtoȱmakeȱsureȱthatȱmyȱchildrenȱhaveȱaccessȱ toȱthisȱfutureȱandȱthatȱIȱdoȱnotȱhaveȱtoȱworryȱaboutȱwhatȱtheyȱmightȱ beȱrunningȱintoȱonȱline.ȱIȱwouldȱlikeȱtoȱkeepȱthatȱoutȱofȱmyȱhouseȱandȱ offȱmyȱcomputer.ȱ ȱ Id.ȱatȱ22,044Ȭ45.ȱLikewise,ȱCongressmanȱWydenȱsaid:ȱ“Weȱareȱallȱagainstȱsmutȱandȱ pornography,ȱand,ȱasȱtheȱparentsȱofȱtwoȱsmallȱcomputerȬliterateȱchildren,ȱmyȱwifeȱ andȱ Iȱ haveȱ seenȱ ourȱ kidsȱ findȱ theirȱ wayȱ intoȱ theseȱ chatȱ roomsȱ thatȱ makeȱ theirȱ middleȬagedȱparentsȱcringe.”ȱId.ȱatȱ22,045.ȱȱ ȱ Asȱbothȱsponsorsȱnoted,ȱtheȱdebateȱbetweenȱtheȱHouseȱandȱtheȱSenateȱwasȱ notȱoverȱtheȱCDA’sȱprimaryȱpurposeȱbutȱratherȱoverȱtheȱbestȱmeansȱtoȱthatȱsharedȱ end.ȱSeeȱid.ȱ(statementȱofȱRep.ȱCox)ȱ(“Howȱshouldȱweȱdoȱthis?ȱ.ȱ.ȱ.ȱMr.ȱChairman,ȱ whatȱweȱwantȱareȱresults.ȱWeȱwantȱtoȱmakeȱsureȱweȱdoȱsomethingȱthatȱactuallyȱ works.”);ȱid.ȱ(statementȱofȱRep.ȱWyden)ȱ(“Soȱletȱusȱallȱstipulateȱrightȱatȱtheȱoutsetȱ theȱimportanceȱofȱprotectingȱourȱkidsȱandȱgoingȱtoȱtheȱissueȱofȱtheȱbestȱwayȱtoȱdoȱ it.”).ȱ Whileȱ theȱ Exonȱ Amendmentȱ wouldȱ haveȱ theȱ FCCȱ regulateȱ onlineȱ obsceneȱ 7ȱ ȱ ȱ materials,ȱtheȱsponsorsȱofȱtheȱHouseȱproposalȱ“believe[d]ȱthatȱparentsȱandȱfamiliesȱ areȱbetterȱsuitedȱtoȱguardȱtheȱportalsȱofȱcyberspaceȱandȱprotectȱourȱchildrenȱthanȱ ourȱGovernmentȱbureaucrats.”ȱId.ȱatȱ22,045ȱ(statementȱofȱRep.ȱWyden).ȱTheyȱalsoȱ fearedȱtheȱeffectsȱtheȱSenate’sȱapproachȱmightȱhaveȱonȱtheȱInternetȱitself.ȱSeeȱid.ȱ (statementȱ ofȱ Rep.ȱ Cox)ȱ (“[Theȱ amendment]ȱ willȱ establishȱ asȱ theȱ policyȱ ofȱ theȱ Unitedȱ Statesȱ thatȱ weȱ doȱ notȱ wishȱ toȱ haveȱ contentȱ regulationȱ byȱ theȱ Federalȱ Governmentȱ ofȱ whatȱ isȱ onȱ theȱ Internet,ȱ thatȱ weȱ doȱ notȱ wishȱ toȱ haveȱ aȱ Federalȱ ComputerȱCommissionȱwithȱanȱarmyȱofȱbureaucratsȱregulatingȱtheȱInternetȱ.ȱ.ȱ.ȱ.”).ȱ TheȱCoxȬWydenȱAmendmentȱthereforeȱsoughtȱtoȱempowerȱinteractiveȱcomputerȱ serviceȱ providersȱ toȱ selfȬregulate,ȱ andȱ toȱ provideȱ toolsȱ forȱ parentsȱ toȱ regulate,ȱ children’sȱaccessȱtoȱinappropriateȱmaterial.ȱSeeȱS.ȱRep.ȱNo.ȱ104Ȭ230,ȱatȱ194ȱ(1996)ȱ (Conf.ȱRep.);ȱ141ȱCong.ȱRec.ȱ22,045ȱ(statementȱofȱRep.ȱCox).ȱ ȱ Thereȱwasȱonlyȱoneȱproblemȱwithȱthisȱapproach,ȱasȱtheȱHouseȱsponsorsȱsawȱ it.ȱAȱNewȱYorkȱStateȱtrialȱcourtȱhadȱrecentlyȱruledȱthatȱtheȱonlineȱserviceȱProdigy,ȱ byȱ decidingȱ toȱ removeȱ certainȱ indecentȱ materialȱ fromȱ itsȱ site,ȱ hadȱ becomeȱ aȱ “publisher”ȱ andȱ thusȱ wasȱ liableȱ forȱ defamationȱ whenȱ itȱ failedȱ toȱ removeȱ otherȱ objectionableȱcontent.ȱStrattonȬOakmont,ȱInc.ȱv.ȱProdigyȱServs.ȱCo.,ȱ1995ȱWLȱ323710,ȱ 8ȱ ȱ ȱ atȱ *4ȱ (N.Y.ȱ Sup.ȱ Ct.ȱ Mayȱ 24,ȱ 1995)ȱ (unpublished).ȱ Theȱ authorsȱ ofȱ §ȱ 230ȱ sawȱ theȱ StrattonȬOakmontȱdecisionȱasȱindicativeȱofȱaȱ“legalȱsystemȱ[that]ȱprovidesȱaȱmassiveȱ disincentiveȱforȱtheȱpeopleȱwhoȱmightȱbestȱhelpȱusȱcontrolȱtheȱInternetȱtoȱdoȱso.”ȱ 141ȱCong.ȱRec.ȱ22,045ȱ(statementȱofȱRep.ȱCox).ȱCoxȬWydenȱwasȱdesigned,ȱinȱlargeȱ part,ȱtoȱremoveȱthatȱdisincentive.ȱSeeȱS.ȱRep.ȱNo.ȱ104Ȭ230,ȱatȱ194.ȱ ȱ TheȱHouseȱhavingȱpassedȱtheȱCoxȬWydenȱAmendmentȱandȱtheȱSenateȱtheȱ ExonȱAmendment,ȱtheȱconferenceȱcommitteeȱhadȱbeforeȱitȱtwoȱalternativeȱvisionsȱ forȱ counteringȱ theȱ spreadȱ ofȱ indecentȱ onlineȱ materialȱ toȱminors.ȱ Theȱcommitteeȱ choseȱnotȱtoȱchoose.ȱCongressȱinsteadȱadoptedȱbothȱamendmentsȱasȱpartȱofȱaȱfinalȱ CommunicationsȱDecencyȱAct.ȱSeeȱTelecommunicationsȱActȱofȱ1996,ȱ§§ȱ502,ȱ509,ȱ 110ȱStat.ȱatȱ133Ȭ39;ȱReno,ȱ521ȱU.S.ȱatȱ858ȱn.24.2ȱTheȱSupremeȱCourtȱpromptlyȱstruckȱ downȱ twoȱ majorȱ provisionsȱ ofȱ theȱ Exonȱ Amendmentȱ asȱ unconstitutionallyȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 2ȱ Itȱ helpedȱ thatȱ theȱ CoxȬWydenȱ Amendmentȱ exemptedȱ fromȱ itsȱ deregulatoryȱ regimeȱ theȱ veryȱ provisionsȱ thatȱ theȱ Exonȱ Amendmentȱ strengthened,ȱ seeȱ TelecommunicationsȱActȱofȱ1996,ȱ§§ȱ502,ȱ507Ȭ508,ȱ509(d)(1),ȱ110ȱStat.ȱatȱ133Ȭ39,ȱandȱthatȱ CongressȱstrippedȱfromȱtheȱHouseȱbillȱaȱprovisionȱthatȱwouldȱhaveȱdeniedȱjurisdictionȱ toȱtheȱFCCȱtoȱregulateȱtheȱInternet,ȱcompareȱid.ȱ§ȱ509,ȱ110ȱStat.ȱatȱ138ȱ(eliminatingȱoriginalȱ ȱ§ȱ509(d)),ȱwithȱ141ȱCong.ȱRec.ȱ22,044ȱ(includingȱoriginalȱ§ȱ509(d)).ȱ 9ȱ ȱ overbroadȱ underȱ theȱ Firstȱ Amendment,ȱ leavingȱ theȱ newȱ §ȱ 230ȱ asȱ theȱ dominantȱ forceȱforȱsecuringȱdecencyȱonȱtheȱInternet.ȱSeeȱReno,ȱ521ȱU.S.ȱatȱ849.ȱ ȱ Sectionȱ 230ȱ overruledȱ StrattonȬOakmontȱ throughȱ twoȱ interlockingȱ provisions,ȱ bothȱ ofȱ whichȱ survivedȱ theȱ legislativeȱ processȱ unscathed.ȱ Theȱ first,ȱ whichȱisȱatȱissueȱinȱthisȱcase,ȱstatesȱthatȱ“[n]oȱproviderȱorȱuserȱofȱanȱinteractiveȱ computerȱserviceȱshallȱbeȱtreatedȱasȱtheȱpublisherȱorȱspeakerȱofȱanyȱinformationȱ providedȱ byȱ anotherȱ informationȱ contentȱ provider.”ȱ 47ȱ U.S.C.ȱ §ȱ 230(c)(1).ȱ Theȱ secondȱ provisionȱ eliminatesȱ liabilityȱ forȱ interactiveȱ computerȱ serviceȱ providersȱ andȱusersȱforȱ“anyȱactionȱvoluntarilyȱtakenȱinȱgoodȱfaithȱtoȱrestrictȱaccessȱtoȱorȱ availabilityȱofȱmaterialȱthatȱtheȱproviderȱorȱuserȱconsidersȱtoȱbeȱ.ȱ.ȱ.ȱobjectionable,”ȱ orȱ“anyȱactionȱtakenȱtoȱenableȱorȱmakeȱavailableȱtoȱ.ȱ.ȱ.ȱothersȱtheȱtechnicalȱmeansȱ toȱrestrictȱaccessȱtoȱ[objectionable]ȱmaterial.”ȱId.ȱ§ȱ230(c)(2).ȱTheseȱtwoȱsubsectionsȱ tackle,ȱ inȱ overlappingȱ fashion,ȱ theȱ twoȱ jurisprudentialȱ movesȱ ofȱ theȱ StrattonȬ Oakmontȱ court:ȱ first,ȱ thatȱ Prodigy’sȱ decisionȱ toȱ screenȱ postsȱ forȱ offensivenessȱ renderedȱ itȱ “aȱ publisherȱ ratherȱ thanȱ aȱ distributor,”ȱ 1995ȱ WLȱ 323710,ȱ atȱ *4;ȱ andȱ second,ȱ thatȱ byȱmakingȱ goodȬfaithȱ effortsȱ toȱ removeȱ offensiveȱ materialȱ Prodigyȱ becameȱliableȱforȱanyȱactionableȱmaterialȱitȱdidȱnotȱremove.ȱ 10ȱ ȱ ȱ ȱ Theȱlegislativeȱhistoryȱillustratesȱthatȱinȱpassingȱ§ȱ230ȱCongressȱwasȱfocusedȱ squarelyȱonȱprotectingȱminorsȱfromȱoffensiveȱonlineȱmaterial,ȱandȱthatȱitȱsoughtȱ toȱ doȱ soȱ byȱ “empoweringȱ parentsȱ toȱ determineȱ theȱ contentȱ ofȱ communicationsȱ theirȱchildrenȱreceiveȱthroughȱinteractiveȱcomputerȱservices.”ȱS.ȱRep.ȱNo.ȱ104Ȭ230,ȱ atȱ 194.ȱ Theȱ “policy”ȱ sectionȱ ofȱ §ȱ230’sȱ textȱ reflectsȱ thisȱ goal.ȱ Seeȱ 47ȱ U.S.C.ȱ §ȱ230(b)(3)Ȭ(4).3ȱ Itȱ isȱ notȱ surprising,ȱ then,ȱ thatȱ Congressȱ emphasizedȱ theȱ narrowȱ civilȱ liabilityȱ shieldȱ thatȱ becameȱ §ȱ230(c)(2),ȱ ratherȱ thanȱ theȱ broadȱ ruleȱ ofȱ constructionȱlaidȱoutȱinȱ§ȱ230(c)(1).ȱIndeed,ȱtheȱconferenceȱcommitteeȱsummarizedȱ §ȱ230ȱbyȱstatingȱthatȱitȱ“providesȱ‘GoodȱSamaritan’ȱprotectionsȱfromȱcivilȱliabilityȱ forȱprovidersȱorȱusersȱofȱanȱinteractiveȱcomputerȱserviceȱforȱactionsȱtoȱrestrictȱorȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 3ȱTheȱpolicyȱsectionȱofȱtheȱstatuteȱalsoȱexpressesȱCongress’sȱdesireȱ“toȱpreserveȱtheȱ vibrantȱ andȱ competitiveȱ freeȱ marketȱ thatȱ presentlyȱ existsȱ forȱ theȱ Internetȱ andȱ otherȱ interactiveȱ computerȱ services,ȱ unfetteredȱ byȱ Federalȱ orȱ Stateȱ regulation.”ȱ 47ȱ U.S.C.ȱ §ȱ230(b)(2).ȱ Itȱ isȱ thereforeȱ trueȱ thatȱ “Sectionȱ 230ȱ wasȱ enacted,ȱ inȱ part,ȱ toȱ maintainȱ theȱ robustȱnatureȱofȱInternetȱcommunication.”ȱRicciȱv.ȱTeamstersȱUnionȱLocalȱ456,ȱ781ȱF.3dȱ25,ȱ 28ȱ(emphasisȱadded)ȱ(quotingȱZeranȱv.ȱAm.ȱOnline,ȱInc.,ȱ129ȱF.3dȱ327,ȱ330ȱ(4thȱCir.ȱ1997));ȱ seeȱ ante,ȱ atȱ 24.ȱ Asȱ theȱ legislativeȱ historyȱ laidȱ outȱ inȱ thisȱ opinionȱ shows,ȱ however,ȱ oneȱ cannotȱ fullyȱ understandȱ theȱ purposeȱ ofȱ §ȱ 230ȱ withoutȱ consideringȱ thatȱ itȱ wasȱ oneȱ chamber’sȱproposalȱinȱaȱdisagreementȱbetweenȱtheȱtwoȱhousesȱofȱCongressȱoverȱhowȱbestȱ toȱ shieldȱ childrenȱ fromȱ indecentȱ material,ȱ andȱ thatȱ inȱ thatȱ contestȱ theȱ Houseȱ wasȱ principallyȱ concernedȱ withȱ twoȱ things:ȱ (1)ȱ overrulingȱ StrattonȬOakmontȱ andȱ (2)ȱpreventingȱ“aȱFederalȱComputerȱCommissionȱwithȱanȱarmyȱofȱbureaucratsȱregulatingȱ ȱtheȱInternet.”ȱ141ȱCong.ȱRec.ȱ22,045ȱ(statementȱofȱRep.ȱCox).ȱ 11ȱ ȱ toȱenableȱrestrictionȱofȱaccessȱtoȱobjectionableȱonlineȱmaterial”—aȱdescriptionȱthatȱ couldȱjustȱasȱeasilyȱhaveȱappliedȱtoȱ§ȱ230(c)(2)ȱalone.ȱS.ȱRep.ȱNo.ȱ104Ȭ230,ȱatȱ194.ȱ Congressȱ alsoȱ titledȱ theȱ entiretyȱ ofȱ §ȱ230(c)ȱ “Protectionȱ forȱ ‘Goodȱ Samaritan’ȱ blockingȱandȱscreeningȱofȱoffensiveȱmaterial,”ȱsuggestingȱthatȱtheȱdefinitionalȱruleȱ outlinedȱ inȱ §ȱ 230(c)(1)ȱ mayȱ haveȱ beenȱ envisionedȱ asȱ supportingȱ orȱ workingȱ inȱ tandemȱwithȱtheȱcivilȱliabilityȱshieldȱinȱ§ȱ230(c)(2).ȱȱ ȱ Noneȱofȱthisȱisȱtoȱsayȱthatȱ§ȱ230(c)(1)ȱexemptsȱinteractiveȱcomputerȱserviceȱ providersȱ fromȱ publisherȱ treatmentȱ onlyȱ whenȱ theyȱ removeȱ indecentȱ content.ȱ Statutoryȱ textȱ cannotȱ beȱ ignored,ȱ andȱ Congressȱ grabbedȱ aȱ bazookaȱ toȱ swatȱ theȱ StrattonȬOakmontȱfly.ȱWhateverȱprototypicalȱsituationȱitsȱdraftersȱmayȱhaveȱhadȱinȱ mind,ȱ §ȱ230(c)(1)ȱ doesȱ notȱ limitȱ itsȱ protectionȱ toȱ situationsȱ involvingȱ “obsceneȱ material”ȱprovidedȱbyȱothers,ȱinsteadȱusingȱtheȱexpansiveȱwordȱ“information.”4ȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 4ȱThisȱpoint—thatȱCongressȱchoseȱbroaderȱlanguageȱthanȱmayȱhaveȱbeenȱnecessaryȱ toȱ accomplishȱ itsȱ primaryȱ goal—shouldȱ notȱ beȱ confusedȱ withȱ theȱ Seventhȱ Circuit’sȱ rationaleȱ forȱ§ȱ230(c)(1)’sȱgeneralȱ application:ȱ thatȱ “aȱ law’sȱ scopeȱ oftenȱ differsȱ fromȱitsȱ genesis.”ȱSeeȱChi.ȱLawyersȱCmte.ȱforȱCivilȱRightsȱUnderȱLaw,ȱInc.ȱv.ȱCraigslist,ȱInc.,ȱ519ȱF.3dȱ 666,ȱ671ȱ(7thȱCir.ȱ2008).ȱTrueȱasȱthisȱaxiomȱmightȱbe,ȱitȱdoesȱnotȱapplyȱhere—theȱlanguageȱ ofȱ §ȱ230(c)(1)ȱ remainedȱ untouchedȱ fromȱ introductionȱ toȱ passage.ȱ Norȱ isȱ thereȱ anyȱ evidenceȱfromȱtheȱlegislativeȱrecordȱthatȱinterestȱgroupsȱalteredȱtheȱstatutoryȱlanguage.ȱ Butȱ cf.ȱ id.ȱ (“Onceȱ theȱ legislativeȱ processȱ getsȱ rolling,ȱ interestȱ groupsȱ seekȱ (andȱ oftenȱ ȱ obtain)ȱotherȱprovisions.”).ȱThatȱ§ȱ230(c)(1)’sȱbreadthȱflowedȱfromȱCongress’sȱdesireȱtoȱ 12ȱ ȱ Illuminatingȱ Congress’sȱ originalȱ intentȱ does,ȱ however,ȱ underscoreȱ theȱ extentȱ ofȱ §ȱ230(c)(1)’sȱsubsequentȱmissionȱcreep.ȱGivenȱhowȱfarȱbothȱFacebook’sȱsuggestionȱ algorithmsȱandȱplaintiffs’ȱterrorismȱclaimsȱswimȱfromȱtheȱshoreȱofȱcongressionalȱ purpose,ȱcautionȱisȱwarrantedȱbeforeȱcourtsȱextendȱtheȱCDA’sȱreachȱanyȱfurther.ȱ II. ȱ WithȱtheȱCDA’sȱbackgroundȱinȱmind,ȱIȱturnȱtoȱtheȱtext.ȱByȱitsȱplainȱterms,ȱ §ȱ230ȱdoesȱnotȱapplyȱwheneverȱaȱclaimȱwouldȱtreatȱtheȱdefendantȱasȱ“aȱpublisher”ȱ inȱtheȱabstract,ȱimmunizingȱdefendantsȱfromȱliabilityȱstemmingȱfromȱanyȱactivityȱ inȱwhichȱoneȱthinksȱpublishingȱcompaniesȱcommonlyȱengage.ȱContraȱante,ȱatȱ30Ȭ 31,ȱ33Ȭ34,ȱ49.ȱItȱstates,ȱmoreȱspecifically,ȱthatȱ“[n]oȱproviderȱorȱuserȱofȱanȱinteractiveȱ computerȱ serviceȱ shallȱ beȱ treatedȱ asȱ theȱ publisherȱ orȱ speakerȱ ofȱ anyȱ informationȱ providedȱbyȱanotherȱinformationȱcontentȱprovider.”ȱ47ȱU.S.C.ȱ§ȱ230(c)(1)ȱ(emphasesȱ added).ȱ “Hereȱ grammarȱ andȱ usageȱ establishȱ thatȱ ‘the’ȱ isȱ aȱ functionȱ wordȱ indicatingȱ thatȱ aȱ followingȱ nounȱ orȱ nounȱ equivalentȱ isȱ definiteȱ .ȱ .ȱ .ȱ .”ȱ Nielsenȱ v.ȱ Preap,ȱ139ȱS.ȱCt.ȱ954,ȱ965ȱ(2019)ȱ(citationȱandȱinternalȱquotationȱmarksȱomitted).ȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ overruleȱStrattonȬOakmont,ȱratherȱthanȱfromȱmereȱinterestȱgroupȱprotectionism,ȱmatters.ȱ 13ȱ ȱ Theȱwordȱ“publisher”ȱinȱthisȱstatuteȱisȱthusȱinextricablyȱlinkedȱtoȱtheȱ“informationȱ providedȱ byȱ another.”ȱ Theȱ questionȱ isȱ whetherȱ aȱ plaintiff’sȱ claimȱ arisesȱ fromȱ aȱ thirdȱparty’sȱinformation,ȱand—crucially—whetherȱtoȱestablishȱtheȱclaimȱtheȱcourtȱ mustȱnecessarilyȱviewȱtheȱdefendant,ȱnotȱasȱaȱpublisherȱinȱtheȱabstract,ȱbutȱratherȱ asȱtheȱpublisherȱofȱthatȱthirdȬpartyȱinformation.ȱSeeȱFTCȱv.ȱLeadClickȱMedia,ȱLLC,ȱ 838ȱF.3dȱ158,ȱ175ȱ(2dȱCir.ȱ2016)ȱ(statingȱinquiryȱasȱ“whetherȱtheȱcauseȱofȱactionȱ inherentlyȱrequiresȱtheȱcourtȱtoȱtreatȱtheȱdefendantȱasȱtheȱ‘publisherȱorȱspeaker’ȱofȱ contentȱprovidedȱbyȱanother”).ȱ Forȱthisȱreason,ȱ§ȱ230(c)(1)ȱdoesȱnotȱnecessarilyȱimmunizeȱdefendantsȱfromȱ claimsȱbasedȱonȱpromotingȱcontentȱorȱsellingȱadvertising,ȱevenȱifȱthoseȱactivitiesȱ mightȱ beȱ commonȱ amongȱ publishingȱ companiesȱ nowadays.ȱ Aȱ publisherȱ mightȱ writeȱanȱemailȱpromotingȱaȱthirdȬpartyȱeventȱtoȱitsȱreaders,ȱforȱexample,ȱbutȱtheȱ publisherȱ wouldȱ beȱ theȱ authorȱ ofȱ theȱ underlyingȱ contentȱ andȱ thereforeȱ notȱ immuneȱ fromȱ suitȱ basedȱ onȱ thatȱ promotion.ȱ Seeȱ 47ȱ U.S.C.ȱ §ȱ 230(c)(1),ȱ (f)(3).ȱ Similarly,ȱtheȱfactȱthatȱpublishersȱmayȱsellȱadvertisingȱbasedȱonȱuserȱdataȱdoesȱnotȱ immunizeȱtheȱpublisherȱifȱsomeoneȱbringsȱaȱclaimȱbasedȱonȱtheȱpublisher’sȱsellingȱ ofȱtheȱdata,ȱbecauseȱtheȱclaimȱwouldȱnotȱtreatȱtheȱdefendantȱasȱtheȱpublisherȱofȱaȱ 14ȱ ȱ ȱ thirdȱ party’sȱ content.ȱ Cf.ȱ Oberdorfȱ v.ȱ Amazon.comȱ Inc.,ȱ No.ȱ 18Ȭ1041,ȱ 2019ȱ WLȱ 2849153,ȱatȱ*12ȱ(3dȱCir.ȱJulyȱ3,ȱ2019)ȱ(holdingȱthatȱtheȱCDAȱdoesȱnotȱbarȱclaimsȱ againstȱAmazon.comȱ“toȱtheȱextentȱthat”ȱtheyȱ“relyȱonȱAmazon’sȱroleȱasȱanȱactorȱ inȱtheȱsalesȱprocess,”ȱincludingȱbothȱ“selling”ȱandȱ“marketing”).ȱSectionȱ230(c)(1)ȱ limitsȱliabilityȱbasedȱonȱtheȱfunctionȱtheȱdefendantȱperforms,ȱnotȱitsȱidentity.ȱ Accordingly,ȱourȱprecedentȱdoesȱnotȱgrantȱpublishersȱCDAȱimmunityȱforȱ theȱfullȱrangeȱofȱactivitiesȱinȱwhichȱtheyȱmightȱengage.ȱRather,ȱitȱ“barsȱlawsuitsȱ seekingȱtoȱholdȱaȱserviceȱproviderȱliableȱforȱitsȱexerciseȱofȱaȱpublisher’sȱtraditionalȱ editorialȱfunctions—suchȱasȱdecidingȱwhetherȱtoȱpublish,ȱwithdraw,ȱpostponeȱorȱ alterȱ content”ȱ providedȱ byȱ anotherȱ forȱ publication.ȱ LeadClick,ȱ 838ȱ F.3dȱ atȱ 174ȱ (citationȱandȱinternalȱquotationȱmarksȱomitted);ȱaccordȱOberdorf,ȱ2019ȱWLȱ2849153,ȱ atȱ*10;ȱJaneȱDoeȱNo.ȱ1ȱv.ȱBackpage.com,ȱLLC,ȱ817ȱF.3dȱ12,ȱ19ȱ(1stȱCir.ȱ2016);ȱJonesȱv.ȱ Dirtyȱ Worldȱ Entm’tȱ Recordingsȱ LLC,ȱ 755ȱ F.3dȱ 398,ȱ 407ȱ (6thȱ Cir.ȱ 2014);ȱ Barnesȱ v.ȱ Yahoo!,ȱInc.,ȱ570ȱF.3dȱ1096,ȱ1102ȱ(9thȱCir.ȱ2009);ȱZeran,ȱ129ȱF.3dȱatȱ330;ȱseeȱKlaymanȱ v.ȱZuckerberg,ȱ753ȱF.3dȱ1354,ȱ1359ȱ(D.C.ȱCir.ȱ2014);ȱBenȱEzra,ȱWeinstein,ȱ&ȱCo.,ȱInc.ȱv.ȱ Am.ȱOnlineȱInc.,ȱ206ȱF.3dȱ980,ȱ986ȱ(10thȱCir.ȱ2000).ȱForȱinstance,ȱaȱclaimȱagainstȱaȱ newspaperȱbasedȱonȱtheȱcontentȱofȱaȱclassifiedȱadȱ(orȱtheȱdecisionȱtoȱpublishȱorȱ 15ȱ ȱ ȱ withdrawȱthatȱad)ȱwouldȱfailȱunderȱtheȱCDAȱnotȱbecauseȱnewspapersȱtraditionallyȱ publishȱclassifiedȱads,ȱbutȱratherȱbecauseȱsuchȱaȱclaimȱwouldȱnecessarilyȱtreatȱtheȱ newspaperȱasȱtheȱpublisherȱofȱtheȱadȬmaker’sȱcontent.ȱSimilarly,ȱtheȱnewspaperȱ doesȱnotȱactȱasȱanȱ“informationȱcontentȱprovider”—andȱthusȱmaintainsȱitsȱCDAȱ protection—whenȱitȱdecidesȱtoȱrunȱaȱclassifiedȱadȱbecauseȱitȱneitherȱ“creates”ȱnorȱ “develops”ȱtheȱinformationȱinȱtheȱad.ȱ47ȱU.S.C.ȱ§ȱ230(f)(3).ȱȱ Thisȱ caseȱ isȱ different.ȱ Lookingȱ beyondȱ Facebook’sȱ “broadȱ statementsȱ ofȱ immunity”ȱandȱrelyingȱ“ratherȱonȱaȱcarefulȱexegesisȱofȱtheȱstatutoryȱlanguage,”ȱ Barnes,ȱ570ȱF.3dȱatȱ1100,ȱtheȱCDAȱdoesȱnotȱprotectȱFacebook’sȱfriendȬȱandȱcontentȬ suggestionȱalgorithms.ȱAȱcombinationȱofȱtwoȱfactors,ȱinȱmyȱview,ȱconfirmsȱthatȱ claimsȱbasedȱonȱtheseȱalgorithmsȱdoȱnotȱinherentlyȱtreatȱFacebookȱasȱtheȱpublisherȱ ofȱ thirdȬpartyȱ content.5ȱ First,ȱ Facebookȱ usesȱ theȱ algorithmsȱ toȱ createȱ andȱ communicateȱitsȱownȱmessage:ȱthatȱitȱthinksȱyou,ȱtheȱreader—you,ȱspecifically— willȱ likeȱ thisȱ content.ȱ Andȱ second,ȱ Facebook’sȱ suggestionsȱ contributeȱ toȱ theȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 5ȱManyȱofȱFacebook’sȱalgorithmsȱmentionedȱinȱtheȱPSAC,ȱsuchȱasȱitsȱthirdȬpartyȱ advertisingȱalgorithm,ȱitsȱalgorithmȱthatȱplacesȱcontentȱinȱaȱuser’sȱnewsfeed,ȱandȱ(basedȱ onȱ theȱ limitedȱ descriptionȱ inȱ theȱ PSAC)ȱ itsȱ videoȱ recommendationȱ algorithm,ȱ remainȱ ȱimmuneȱunderȱtheȱanalysisȱIȱsetȱoutȱhere.ȱ 16ȱ ȱ creationȱofȱrealȬworldȱsocialȱnetworks.ȱTheȱresultȱofȱatȱleastȱsomeȱsuggestionsȱisȱ notȱ justȱ thatȱ theȱ userȱ consumesȱ aȱ thirdȱ party’sȱ content.ȱ Sometimes,ȱ Facebook’sȱ suggestionsȱallegedlyȱleadȱtheȱuserȱtoȱbecomeȱpartȱofȱaȱuniqueȱglobalȱcommunity,ȱ theȱcreationȱandȱmaintenanceȱofȱwhichȱgoesȱfarȱbeyondȱandȱdiffersȱinȱkindȱfromȱ traditionalȱeditorialȱfunctions.ȱ Itȱisȱtrue,ȱasȱtheȱmajorityȱnotes,ȱseeȱante,ȱatȱ47,ȱthatȱFacebook’sȱalgorithmsȱ relyȱ onȱ andȱ displayȱ users’ȱ content.ȱ However,ȱ thisȱ isȱ notȱ enoughȱ toȱ triggerȱ theȱ protectionsȱofȱ§ȱ230(c)(1).ȱTheȱCDAȱdoesȱnotȱmandateȱ“aȱ‘butȬfor’ȱtestȱthatȱwouldȱ provideȱimmunityȱ.ȱ.ȱ.ȱsolelyȱbecauseȱaȱcauseȱofȱactionȱwouldȱnotȱotherwiseȱhaveȱ accruedȱ butȱ forȱ theȱ thirdȬpartyȱ content.”ȱ HomeAway.com,ȱ Inc.ȱ v.ȱ Cityȱ ofȱ Santaȱ Monica,ȱ918ȱF.3dȱ676,ȱ682ȱ(9thȱCir.ȱ2019).ȱRather,ȱtoȱfallȱwithinȱ§ȱ230(c)(1)’sȱradius,ȱ theȱclaimȱatȱissueȱmustȱinherentlyȱfaultȱtheȱdefendant’sȱactivityȱasȱtheȱpublisherȱofȱ specificȱ thirdȬpartyȱ content.ȱ Plaintiffs’ȱ claimsȱ aboutȱ Facebook’sȱ suggestionȱ algorithmsȱdoȱnotȱdoȱthis.ȱTheȱcomplaintȱallegesȱthatȱ“Facebookȱcollectsȱdetailedȱ informationȱ aboutȱ itsȱ users,ȱ including,ȱ interȱ alia,ȱ theȱ contentȱ theyȱ post,ȱ typeȱ ofȱ contentȱ theyȱ viewȱ orȱ engageȱ with,ȱ peopleȱ theyȱ communicateȱ with,ȱ groupsȱ theyȱ belongȱtoȱandȱhowȱtheyȱinteractȱwithȱsuchȱgroups,ȱvisitsȱtoȱthirdȱpartyȱwebsites,ȱ 17ȱ ȱ ȱ appsȱandȱFacebookȱpartners.”ȱApp’xȱ345ȱ¶ȱ608.ȱThenȱtheȱalgorithmsȱ“utilizeȱtheȱ collectedȱdataȱtoȱsuggestȱfriends,ȱgroups,ȱproducts,ȱservicesȱandȱlocalȱevents,ȱandȱ targetȱads”ȱbasedȱonȱeachȱuser’sȱinput.ȱId.ȱatȱ346ȱ¶ȱ610.ȱȱ Ifȱ aȱ thirdȱ partyȱ gotȱ accessȱ toȱ Facebookȱ users’ȱ data,ȱ analyzedȱ itȱ usingȱ aȱ proprietaryȱalgorithm,ȱandȱsentȱitsȱownȱmessagesȱtoȱFacebookȱusersȱsuggestingȱ thatȱpeopleȱbecomeȱfriendsȱorȱattendȱoneȱanother’sȱevents,ȱtheȱthirdȱpartyȱwouldȱ notȱbeȱprotectedȱasȱ“theȱpublisher”ȱofȱtheȱusers’ȱinformation.ȱSimilarly,ȱifȱFacebookȱ wereȱtoȱuseȱtheȱalgorithmsȱtoȱtargetȱitsȱownȱmaterialȱtoȱparticularȱusers,ȱsuchȱthatȱ theȱresultingȱpostsȱconsistedȱofȱ“informationȱprovidedȱby”ȱFacebookȱratherȱthanȱ byȱ “anotherȱ informationȱ contentȱ provider,”ȱ §ȱ230(c)(1),ȱ Facebookȱ clearlyȱ wouldȱ notȱbeȱimmuneȱforȱthatȱindependentȱmessage.ȱ Yetȱ thatȱ isȱ ultimatelyȱ whatȱ plaintiffsȱ allegeȱ Facebookȱ isȱ doing.ȱ Theȱ PSACȱ allegesȱthatȱFacebookȱ“activelyȱprovidesȱ‘friendȱsuggestions’ȱbetweenȱusersȱwhoȱ haveȱexpressedȱsimilarȱinterests,”ȱandȱthatȱitȱ“activelyȱsuggestsȱgroupsȱandȱeventsȱ toȱusers.”ȱApp’xȱ346ȱ¶¶ȱ612Ȭ13.ȱFacebook’sȱalgorithmsȱthusȱallegedlyȱprovideȱtheȱ userȱwithȱaȱmessageȱfromȱFacebook.ȱFacebookȱisȱtellingȱusers—perhapsȱimplicitly,ȱ butȱclearly—thatȱtheyȱwouldȱlikeȱtheseȱpeople,ȱgroups,ȱorȱevents.ȱInȱthisȱrespect,ȱ 18ȱ ȱ ȱ Facebookȱ“doesȱnotȱmerelyȱprovideȱaȱframeworkȱthatȱcouldȱbeȱutilizedȱforȱproperȱ orȱimproperȱpurposes;ȱrather,ȱ[Facebook’s]ȱworkȱinȱdeveloping”ȱtheȱalgorithmȱandȱ suggestingȱ connectionsȱ toȱ usersȱ basedȱ onȱ theirȱ priorȱ activityȱ onȱ Facebook,ȱ includingȱ theirȱ sharedȱ interestȱ inȱ terrorism,ȱ “isȱ directlyȱ relatedȱ toȱ theȱ allegedȱ illegalityȱofȱtheȱsite.”ȱFairȱHousingȱCouncilȱofȱSanȱFernandoȱValleyȱv.ȱRoommates.Com,ȱ LLC,ȱ 521ȱ F.3dȱ 1157,ȱ 1171ȱ (9thȱ Cir.ȱ 2008)ȱ (enȱ banc).ȱ Theȱ factȱ thatȱ Facebookȱ alsoȱ publishesȱthirdȬpartyȱcontentȱshouldȱnotȱcauseȱusȱtoȱconflateȱitsȱtwoȱseparateȱrolesȱ withȱrespectȱtoȱitsȱusersȱandȱtheirȱinformation.ȱFacebookȱmayȱbeȱimmuneȱunderȱ theȱ CDAȱ fromȱ plaintiffs’ȱ challengeȱ toȱ itsȱ allowanceȱ ofȱ Hamasȱ accounts,ȱ sinceȱ FacebookȱactsȱsolelyȱasȱtheȱpublisherȱofȱtheȱHamasȱusers’ȱcontent.ȱThatȱdoesȱnotȱ mean,ȱthough,ȱthatȱitȱisȱalsoȱimmuneȱwhenȱitȱconductsȱstatisticalȱanalysesȱofȱthatȱ informationȱandȱdeliversȱaȱmessageȱbasedȱonȱthoseȱanalyses.ȱ Moreover,ȱinȱpartȱthroughȱitsȱuseȱofȱfriend,ȱgroup,ȱandȱeventȱsuggestions,ȱ Facebookȱ isȱ doingȱ moreȱ thanȱ justȱ publishingȱ content:ȱ itȱ isȱ proactivelyȱ creatingȱ networksȱofȱpeople.ȱItsȱalgorithmsȱforgeȱrealȬworldȱ(ifȱdigital)ȱconnectionsȱthroughȱ friendȱandȱgroupȱsuggestions,ȱandȱtheyȱattemptȱtoȱcreateȱsimilarȱconnectionsȱinȱ theȱ physicalȱ worldȱ throughȱ eventȱ suggestions.ȱ Theȱ cumulativeȱ effectȱ ofȱ 19ȱ ȱ ȱ recommendingȱseveralȱfriends,ȱorȱseveralȱgroupsȱorȱevents,ȱhasȱanȱimpactȱgreaterȱ thanȱtheȱsumȱofȱeachȱsuggestion.ȱItȱenvelopsȱtheȱuser,ȱimmersingȱherȱinȱanȱentireȱ universeȱfilledȱwithȱpeople,ȱideas,ȱandȱeventsȱsheȱmayȱneverȱhaveȱdiscoveredȱonȱ herȱ own.ȱ Accordingȱ toȱ theȱ allegationsȱ inȱ theȱ complaint,ȱ Facebookȱ designedȱ itsȱ websiteȱ forȱ thisȱ veryȱ purpose.ȱ “Facebookȱ hasȱ describedȱ itselfȱ asȱ aȱ providerȱ ofȱ productsȱandȱservicesȱthatȱenableȱusersȱ.ȱ.ȱ.ȱtoȱfindȱandȱconnectȱwithȱotherȱusersȱ .ȱ.ȱ.ȱ.”ȱApp’xȱ250ȱ¶ȱ129.ȱCEOȱMarkȱZuckerbergȱhasȱsimilarlyȱdescribedȱFacebookȱasȱ “build[ing]ȱ toolsȱ toȱ helpȱ peopleȱ connectȱ withȱ theȱ peopleȱ theyȱ want,”ȱ therebyȱ “extendingȱpeople’sȱcapacityȱtoȱbuildȱandȱmaintainȱrelationships.”ȱId.ȱatȱ251ȱ¶ȱ132.ȱ Ofȱcourse,ȱFacebookȱisȱnotȱtheȱonlyȱcompanyȱthatȱtriesȱtoȱbringȱpeopleȱtogetherȱ thisȱ way,ȱ andȱ perhapsȱ otherȱ publishersȱ tryȱ toȱ introduceȱ theirȱ readersȱ toȱ oneȱ another.ȱ Yetȱ theȱ creationȱ ofȱ socialȱ networksȱ goesȱ farȱ beyondȱ theȱ traditionalȱ editorialȱfunctionsȱthatȱtheȱCDAȱimmunizes.ȱ AnotherȱwayȱtoȱconsiderȱtheȱCDAȱimmunityȱquestionȱisȱtoȱ“lookȱ.ȱ.ȱ.ȱtoȱwhatȱ theȱ dutyȱ atȱ issueȱ actuallyȱ requires:ȱ specifically,ȱ whetherȱ theȱ dutyȱ wouldȱ necessarilyȱrequireȱanȱinternetȱcompanyȱtoȱmonitor[,ȱalter,ȱorȱremove]ȱthirdȬpartyȱ content.”ȱ HomeAway.com,ȱ 918ȱ F.3dȱ atȱ 682.ȱ Here,ȱ too,ȱ theȱ claimsȱ regardingȱ theȱ 20ȱ ȱ ȱ algorithmsȱareȱaȱpoorȱfitȱforȱstatutoryȱimmunity.ȱTheȱdutyȱnotȱtoȱprovideȱmaterialȱ supportȱ toȱ terrorism,ȱ asȱ appliedȱ toȱ Facebook’sȱ useȱ ofȱ theȱ algorithms,ȱ simplyȱ requiresȱ thatȱ Facebookȱ notȱ activelyȱ useȱ thatȱ materialȱ toȱ determineȱ whichȱ ofȱ itsȱ usersȱtoȱconnectȱtoȱeachȱother.ȱItȱcouldȱstopȱusingȱtheȱalgorithmsȱaltogether,ȱforȱ instance.ȱ Or,ȱ shortȱ ofȱ that,ȱ Facebookȱ couldȱ modifyȱ itsȱ algorithmsȱ toȱ stopȱ themȱ introducingȱterroristsȱtoȱoneȱanother.ȱNoneȱofȱthisȱwouldȱchangeȱanyȱunderlyingȱ content,ȱ norȱ wouldȱ itȱ necessarilyȱ requireȱ courtsȱ toȱ assessȱ furtherȱ theȱ difficultȱ questionȱofȱwhetherȱthereȱisȱanȱaffirmativeȱobligationȱtoȱmonitorȱthatȱcontent.ȱ Inȱreachingȱthisȱconclusion,ȱIȱnoteȱthatȱATAȱtortsȱareȱatypical.ȱMostȱofȱtheȱ commonȱtortsȱthatȱmightȱbeȱpleadedȱinȱrelationȱtoȱFacebook’sȱalgorithmsȱ“deriveȱ liabilityȱfromȱbehaviorȱthatȱisȱidenticalȱtoȱpublishingȱorȱspeaking”—forȱinstance,ȱ “publishingȱ defamatoryȱ material;ȱ publishingȱ materialȱ thatȱ inflictsȱ emotionalȱ distress;ȱ orȱ .ȱ .ȱ .ȱ attemptingȱ toȱ deȬpublishȱ hurtfulȱ materialȱ butȱ doingȱ itȱ badly.”ȱ Barnes,ȱ 570ȱ F.3dȱ atȱ 1107.ȱ Theȱ factȱ thatȱ Facebookȱ hasȱ figuredȱ outȱ howȱ toȱ targetȱ materialȱtoȱpeopleȱmoreȱlikelyȱtoȱreadȱitȱdoesȱnotȱmatterȱtoȱaȱdefamationȱclaim,ȱforȱ instance,ȱbecauseȱtheȱmereȱactȱofȱpublishingȱinȱtheȱfirstȱplaceȱcreatesȱliability.ȱ 21ȱ ȱ ȱ Theȱ ATAȱ worksȱ differently.ȱ Plaintiffs’ȱ materialȱ supportȱ andȱ aidingȱ andȱ abettingȱclaimsȱpremiseȱliability,ȱnotȱonȱpublishingȱquaȱpublishing,ȱbutȱratherȱonȱ Facebook’sȱprovisionȱofȱservicesȱandȱpersonnelȱtoȱHamas.ȱItȱhappensȱthatȱtheȱwayȱ inȱ whichȱ Facebookȱ providesȱ theseȱ benefitsȱ includesȱ republishingȱ content,ȱ butȱ Facebook’sȱ dutiesȱ underȱ theȱ ATAȱ ariseȱ separatelyȱ fromȱ theȱ republicationȱ ofȱ content.ȱ Cf.ȱ id.ȱ (determiningȱ thatȱ liabilityȱ onȱ aȱ promissoryȱ estoppelȱ theoryȱ forȱ promisingȱtoȱremoveȱcontentȱ“wouldȱcomeȱnotȱfromȱYahoo’sȱpublishingȱconduct,ȱ butȱfromȱYahoo’sȱmanifestȱintentionȱtoȱbeȱlegallyȱobligatedȱtoȱdoȱsomething,ȱwhichȱ happensȱtoȱbeȱremovalȱofȱmaterialȱfromȱpublication”).ȱForȱinstance,ȱtheȱoperationȱ ofȱtheȱalgorithmsȱisȱallegedlyȱprovisionȱofȱ“expertȱadviceȱorȱassistance,”ȱandȱtheȱ messageȱimpliedȱbyȱFacebook’sȱproddingȱisȱallegedlyȱaȱ“service”ȱorȱanȱattemptȱtoȱ provideȱ“personnel.”ȱ18ȱU.S.C.ȱ§ȱ2339A(b).ȱ Forȱtheseȱreasons,ȱ§ȱ230(c)(1)ȱdoesȱnotȱbarȱplaintiffs’ȱclaims.ȱ III. ȱ Evenȱifȱweȱsentȱthisȱcaseȱbackȱtoȱtheȱdistrictȱcourt,ȱasȱIȱbelieveȱtoȱbeȱtheȱrightȱ course,ȱ theseȱ plaintiffsȱ mightȱ haveȱ provenȱ unableȱ toȱ allegeȱ thatȱ Facebook’sȱ matchmakingȱalgorithmsȱplayedȱaȱroleȱinȱtheȱattacksȱthatȱharmedȱthem.ȱHowever,ȱ 22ȱ ȱ ȱ assumingȱarguendoȱthatȱsuchȱmightȱhaveȱbeenȱtheȱsituationȱhere,ȱIȱdoȱnotȱthinkȱweȱ shouldȱ forecloseȱ theȱ possibilityȱ ofȱ reliefȱ inȱ futureȱ casesȱ ifȱ victimsȱ canȱ plausiblyȱ allegeȱ thatȱ aȱ websiteȱ knowinglyȱ broughtȱ terroristsȱ togetherȱ andȱ thatȱ anȱ attackȱ occurredȱasȱaȱdirectȱresultȱofȱtheȱsite’sȱactions.ȱThoughȱtheȱmajorityȱshutsȱtheȱdoorȱ onȱsuchȱclaims,ȱtoday’sȱdecisionȱalsoȱillustratesȱtheȱextensiveȱimmunityȱthatȱtheȱ currentȱ formulationȱ ofȱ theȱ CDAȱ alreadyȱ extendsȱ toȱ socialȱ mediaȱ companiesȱ forȱ activitiesȱthatȱwereȱundreamtȱofȱinȱ1996.ȱItȱthereforeȱmayȱbeȱtimeȱforȱCongressȱtoȱ reconsiderȱtheȱscopeȱofȱ§ȱ230.ȱ Asȱisȱsoȱoftenȱtheȱcaseȱwithȱnewȱtechnologies,ȱtheȱveryȱqualitiesȱthatȱdriveȱ socialȱ media’sȱ success—itsȱ easeȱ ofȱ use,ȱ openȱ access,ȱ andȱ abilityȱ toȱ connectȱ theȱ world—haveȱ alsoȱ spawnedȱ itsȱ demons.ȱ Plaintiffs’ȱ complaintȱ illustratesȱ howȱ pervasiveȱ andȱ blatantȱ aȱ presenceȱ Hamasȱ andȱ itsȱ leadersȱ haveȱ maintainedȱ onȱ Facebook.ȱHamasȱisȱfarȱfromȱalone—Hezbollah,ȱBokoȱHaram,ȱtheȱRevolutionaryȱ ArmedȱForcesȱofȱColombia,ȱandȱmanyȱotherȱdesignatedȱterroristȱorganizationsȱuseȱ Facebookȱtoȱrecruitȱandȱrouseȱsupporters.ȱVernonȱSilverȱ&ȱSarahȱFrier,ȱTerroristsȱ Areȱ Stillȱ Recruitingȱ onȱ Facebook,ȱ Despiteȱ Zuckerberg’sȱ Reassurances,ȱ Bloombergȱ Businessweekȱ(Mayȱ10,ȱ2018),ȱhttp://www.bloomberg.com/news/articles/2018Ȭ05Ȭ 23ȱ ȱ ȱ 10/terroristsȬcreepȬontoȬfacebookȬasȬfastȬasȬitȬcanȬshutȬthemȬdown.ȱ Recentȱ newsȱ reportsȱ suggestȱ thatȱ manyȱ socialȱ mediaȱ sitesȱ haveȱ beenȱ slowȱ toȱ removeȱ theȱ plethoraȱofȱterroristȱandȱextremistȱaccountsȱpopulatingȱtheirȱplatforms,6ȱandȱthatȱ suchȱ efforts,ȱ whenȱ theyȱ occur,ȱ areȱ oftenȱ underinclusive.ȱ Twitter,ȱ forȱ instance,ȱ bannedȱtheȱKuȱKluxȱKlanȱinȱ2018ȱbutȱallowedȱDavidȱDukeȱtoȱmaintainȱhisȱaccount,ȱ seeȱRooseȱ&ȱConger,ȱsupra,ȱwhileȱresearchersȱfoundȱthatȱFacebookȱremovedȱfewerȱ thanȱhalfȱtheȱterroristȱaccountsȱandȱpostsȱthoseȱresearchersȱidentified,ȱseeȱWatersȱ &ȱPostings,ȱsupra,ȱatȱ8;ȱDesmondȱButlerȱ&ȱBarbaraȱOrtulay,ȱFacebookȱAutoȬGeneratesȱ Videosȱ Celebratingȱ Extremistȱ Images,ȱ Assoc.ȱ Pressȱ (Mayȱ 9,ȱ 2019),ȱ http://apnews.com/f97c24dab4f34bd0b48b36f2988952a4.ȱ Thoseȱ whoseȱ accountsȱ areȱremovedȱoftenȱpopȱupȱagainȱunderȱdifferentȱnamesȱorȱwithȱslightlyȱdifferentȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 6ȱSee,ȱe.g.,ȱGregoryȱWatersȱ&ȱRobertȱPostings,ȱSpidersȱofȱtheȱCaliphate:ȱMappingȱtheȱ Islamicȱ State’sȱ Globalȱ Supportȱ Networkȱ onȱ Facebookȱ 8,ȱ Counterȱ Extremismȱ Projectȱ (Mayȱ 2018),ȱ http://www.counterextremism.com/sites/default/files/Spiders%20of%20the%20ȱ Caliphate%20%28May%202018%29.pdf;ȱYaacovȱBenmelehȱ&ȱFeliceȱMaranz,ȱIsraelȱWarnsȱ Twitterȱ ofȱ Legalȱ Actionȱ Overȱ Requestsȱ toȱ Removeȱ Content,ȱ Bloombergȱ (Mar.ȱ 20,ȱ 2018),ȱ http://www.bloomberg.com/news/articles/2018Ȭ03Ȭ20/israelȬwarnsȬtwitterȬofȬlegalȬ stepsȬoverȬincitementȬtoȬterrorism;ȱ Mikeȱ Isaac,ȱ Twitterȱ Stepsȱ Upȱ Effortsȱ toȱ Thwartȱ Terrorists’ȱ Tweets,ȱ N.Y.ȱ Timesȱ (Feb.ȱ 5,ȱ 2016),ȱ http://www.nytimes.com/2016/02/06/ȱ technology/twitterȬaccountȬsuspensionsȬterrorism.html;ȱ Kevinȱ Rooseȱ &ȱ Kateȱ Conger,ȱ YouTubeȱtoȱRemoveȱThousandsȱofȱVideosȱPushingȱExtremeȱViews,ȱN.Y.ȱTimesȱ(Juneȱ5,ȱ2019),ȱ ȱhttp://www.nytimes.com/2019/06/05/business/youtubeȬremoveȬextremistȬvideos.html.ȱ 24ȱ ȱ languageȱinȱtheirȱprofiles,ȱplayingȱaȱperverseȱandȱdeadlyȱgameȱofȱWhackȬaȬMoleȱ withȱSiliconȱValley.ȱSeeȱIsaac,ȱsupra;ȱSilverȱ&ȱFrier,ȱsupra.ȱ Ofȱcourse,ȱtheȱfailureȱtoȱremoveȱterroristȱcontent,ȱwhileȱanȱimportantȱpolicyȱ concern,ȱ isȱ immunizedȱ underȱ §ȱ 230ȱ asȱ currentlyȱ written.ȱ Untilȱ today,ȱ theȱ sameȱ couldȱnotȱhaveȱbeenȱsaidȱforȱsocialȱmedia’sȱunsolicited,ȱalgorithmicȱspreadingȱofȱ terrorism.ȱ Shieldingȱ internetȱ companiesȱ thatȱ bringȱ terroristsȱ togetherȱ usingȱ algorithmsȱcouldȱleaveȱdangerousȱactivityȱunchecked.ȱȱ TakeȱFacebook.ȱȱAsȱplaintiffsȱallege,ȱitsȱfriendȬsuggestionȱalgorithmȱappearsȱ toȱconnectȱterroristȱsympathizersȱwithȱpinpointȱprecision.ȱForȱinstance,ȱwhileȱtwoȱ researchersȱwereȱstudyingȱIslamicȱStateȱ(“IS”)ȱactivityȱonȱFacebook,ȱoneȱ“receivedȱ dozensȱofȱproȬISȱaccountsȱasȱrecommendedȱfriendsȱafterȱfriendingȱjustȱoneȱproȬISȱ account.”ȱWatersȱ&ȱPostings,ȱsupra,ȱatȱ78.ȱMoreȱdisturbingly,ȱtheȱotherȱ“receivedȱ anȱinfluxȱofȱPhilippinesȬbasedȱISȱsupportersȱandȱfightersȱasȱrecommendedȱfriendsȱ afterȱlikingȱseveralȱnonȬextremistȱnewsȱpagesȱaboutȱMarawiȱandȱtheȱPhilippinesȱ duringȱ IS’sȱ captureȱ ofȱ theȱ city.”ȱ Id.ȱ Newsȱ reportsȱ indicateȱ thatȱ theȱ friendȬ suggestionȱfeatureȱhasȱintroducedȱthousandsȱofȱISȱsympathizersȱtoȱoneȱanother.ȱ SeeȱMartinȱEvans,ȱFacebookȱAccusedȱofȱIntroducingȱExtremistsȱtoȱOneȱAnotherȱThroughȱ 25ȱ ȱ ȱ ‘Suggestedȱ Friends’ȱ Feature,ȱ Theȱ Telegraphȱ (Mayȱ 5,ȱ 2018),ȱ http://www.telegraph.co.uk/news/2018/05/05/facebookȬaccusedȬintroducingȬ extremistsȬoneȬanotherȬsuggested.ȱ Andȱ thisȱ isȱ farȱ fromȱ theȱ onlyȱ Facebookȱ algorithmȱ thatȱ mayȱ steerȱ peopleȱ towardȱterrorism.ȱAnotherȱturnsȱusers’ȱdeclaredȱinterestsȱintoȱaudienceȱcategoriesȱ toȱenableȱmicrotargetedȱadvertising.ȱInȱ2017,ȱactingȱonȱaȱtip,ȱProPublicaȱsoughtȱtoȱ directȱ anȱ adȱ atȱ theȱ algorithmicallyȬcreatedȱ categoryȱ “Jewȱ hater”—whichȱ turnedȱ outȱ toȱ beȱ real,ȱ asȱ wereȱ “Germanȱ Schutzstaffel,”ȱ “Naziȱ Party,”ȱ andȱ “Hitlerȱ didȱ nothingȱ wrong.”ȱ Juliaȱ Angwinȱ etȱ al.,ȱ Facebookȱ Enabledȱ Advertisesȱ toȱ Reachȱ ‘Jewȱ Haters,’ȱProPublicaȱ(Sept.ȱ14,ȱ2017),ȱhttps://www.propublica.org/article/facebookȬ enabledȬadvertisersȬtoȬreachȬjewȬhaters.ȱ Asȱ theȱ “Jewȱ hater”ȱ categoryȱ wasȱ tooȱ smallȱ forȱ Facebookȱ toȱ runȱ anȱ adȱ campaign,ȱ “Facebook’sȱ automatedȱ systemȱ suggestedȱ‘SecondȱAmendment’ȱasȱanȱadditionalȱcategoryȱ.ȱ.ȱ.ȱpresumablyȱbecauseȱ itsȱsystemȱhadȱcorrelatedȱgunȱenthusiastsȱwithȱantiȬSemites.”ȱId.ȱȱ That’sȱnotȱall.ȱAnotherȱFacebookȱalgorithmȱautoȬgeneratesȱbusinessȱpagesȱ byȱ scrapingȱ employmentȱ informationȱ fromȱ users’ȱ profiles;ȱ otherȱ usersȱ canȱ thenȱ “like”ȱtheseȱpages,ȱfollowȱtheirȱposts,ȱandȱseeȱwhoȱelseȱhasȱlikedȱthem.ȱButlerȱ&ȱ 26ȱ ȱ ȱ Ortutay,ȱsupra.ȱProPublicaȱreportsȱthatȱextremistȱorganizationsȱincludingȱalȬQaida,ȱ alȬShabab,ȱ andȱ ISȱ haveȱ suchȱ autoȬcreatedȱ pages,ȱ allowingȱ themȱ toȱ recruitȱ theȱ pages’ȱfollowers.ȱId.ȱTheȱpageȱforȱalȬQaidaȱinȱtheȱArabianȱPeninsulaȱincludedȱtheȱ group’sȱ Wikipediaȱ entryȱ andȱ aȱ propagandaȱ photoȱ ofȱ theȱ damagedȱ USSȱ Cole,ȱ whichȱtheȱgroupȱhadȱbombedȱinȱ2000.ȱId.ȱMeanwhile,ȱaȱfourthȱalgorithmȱintegratesȱ users’ȱphotosȱandȱotherȱmediaȱtoȱgenerateȱvideosȱcommemoratingȱtheirȱpreviousȱ year.ȱId.ȱMilitantsȱgetȱaȱreadyȬmadeȱpropagandaȱclip,ȱcompleteȱwithȱaȱthankȬyouȱ messageȱfromȱFacebook.ȱId.ȱ Thisȱcase,ȱandȱourȱCDAȱanalysis,ȱhasȱcenteredȱonȱtheȱuseȱofȱalgorithmsȱtoȱ fomentȱterrorism.ȱYetȱtheȱconsequencesȱofȱaȱCDAȬdriven,ȱhandsȬoffȱapproachȱtoȱ socialȱ mediaȱ extendȱ muchȱ further.ȱ Socialȱ mediaȱ canȱ beȱ usedȱ byȱ foreignȱ governmentsȱtoȱinterfereȱinȱAmericanȱelections.ȱForȱexample,ȱJusticeȱDepartmentȱ prosecutorsȱ recentlyȱ concludedȱ thatȱ Russianȱ intelligenceȱ agentsȱ createdȱ falseȱ Facebookȱ groupsȱ andȱ accountsȱ inȱ theȱ yearsȱ leadingȱ upȱ toȱ theȱ 2016ȱ electionȱ campaign,ȱbootstrappingȱFacebook’sȱalgorithmȱtoȱspewȱpropagandaȱthatȱreachedȱ betweenȱ29ȱmillionȱandȱ126ȱmillionȱAmericans.ȱSeeȱ1ȱRobertȱS.ȱMuellerȱIII,ȱSpecialȱ Counsel,ȱReportȱonȱtheȱInvestigationȱIntoȱRussianȱInterferenceȱinȱtheȱ2016ȱPresidentialȱ 27ȱ ȱ ȱ Electionȱ24Ȭ26,ȱU.S.ȱDep’tȱofȱJusticeȱ(Marchȱ2019),ȱhttp://www.justice.gov/storage/ȱ report.pdf.ȱ Russiaȱ alsoȱ purchasedȱ overȱ 3,500ȱ advertisementsȱ onȱ Facebookȱ toȱ publicizeȱtheirȱfakeȱFacebookȱgroups,ȱseveralȱofȱwhichȱgrewȱtoȱhaveȱhundredsȱofȱ thousandsȱofȱfollowers.ȱId.ȱatȱ25Ȭ26.ȱOnȱTwitter,ȱRussiaȱdevelopedȱfalseȱaccountsȱ thatȱ impersonatedȱ Americanȱ peopleȱ orȱ groupsȱ andȱ issuedȱ contentȱ designedȱ toȱ influenceȱtheȱelection;ȱitȱthenȱcreatedȱthousandsȱofȱautomatedȱ“bot”ȱaccountsȱtoȱ amplifyȱ theȱ shamȱ Americans’ȱ messages.ȱ Id.ȱ atȱ 26Ȭ28.ȱ Oneȱ fakeȱ accountȱ receivedȱ overȱsixȱmillionȱretweets,ȱtheȱvastȱmajorityȱofȱwhichȱappearȱtoȱhaveȱcomeȱfromȱ realȱ Twitterȱ users.ȱ Seeȱ Gillianȱ Cleary,ȱ Twitterbots:ȱ Anatomyȱ ofȱ aȱ Propagandaȱ Campaign,ȱ Symantecȱ (Juneȱ 5,ȱ 2019),ȱ http://www.symantec.com/blogs/threatȬ intelligence/twitterbotsȬpropagandaȬdisinformation.ȱ Russianȱ intelligenceȱ alsoȱ harnessedȱtheȱreachȱthatȱsocialȱmediaȱgaveȱitsȱfalseȱidentitiesȱtoȱorganizeȱ“dozensȱ ofȱ U.S.ȱ rallies,”ȱ someȱ ofȱ whichȱ “drewȱ hundreds”ȱ ofȱ realȬworldȱ Americans.ȱ Mueller,ȱReport,ȱsupra,ȱatȱ29.ȱRussiaȱcouldȱdoȱallȱthisȱonlyȱbecauseȱsocialȱmediaȱisȱ designedȱtoȱtargetȱmessagesȱlikeȱRussia’sȱtoȱtheȱusersȱmostȱsusceptibleȱtoȱthem.ȱ Whileȱ Russia’sȱ interferenceȱ inȱ theȱ 2016ȱ electionȱ isȱ theȱ bestȬdocumentedȱ exampleȱofȱforeignȱmeddlingȱthroughȱsocialȱmedia,ȱitȱisȱnotȱtheȱonlyȱone.ȱFederalȱ 28ȱ ȱ ȱ intelligenceȱ agenciesȱ expressedȱ concernȱ inȱ theȱ weeksȱ beforeȱ theȱ 2018ȱ midtermȱ electionȱ “aboutȱ ongoingȱ campaignsȱ byȱ Russia,ȱ Chinaȱ andȱ otherȱ foreignȱ actors,ȱ includingȱIran,”ȱtoȱ“influenceȱpublicȱsentiment”ȱthroughȱmeansȱ“includingȱusingȱ socialȱ mediaȱ toȱ amplifyȱ divisiveȱ issues.”ȱ Pressȱ Release,ȱ Officeȱ ofȱ Dir.ȱ ofȱ Nat’lȱ Intelligence,ȱ Jointȱ Statementȱ fromȱ theȱ ODNI,ȱ DOJ,ȱ FBI,ȱ andȱ DHS:ȱ Combattingȱ ForeignȱInfluenceȱinȱU.S.ȱElections,ȱ(Oct.ȱ19,ȱ2018),ȱhttps://www.dni.gov/index.ȱ php/newsroom/pressȬreleases/item/1915ȬjointȬstatementȬfromȬtheȬodniȬdojȬfbiȬ andȬdhsȬcombatingȬforeignȬinfluenceȬinȬuȬsȬelections.ȱNewsȱreportsȱalsoȱsuggestȱ thatȱ Chinaȱ targetsȱ stateȬsponsoredȱ propagandaȱ toȱ Americansȱ onȱ Facebookȱ andȱ purchasesȱ Facebookȱ adsȱ toȱ amplifyȱ itsȱ communications.ȱ Seeȱ Paulȱ Mozur,ȱ Chinaȱ SpreadsȱPropagandaȱtoȱU.S.ȱonȱFacebook,ȱaȱPlatformȱItȱBansȱatȱHome,ȱN.Y.ȱTimesȱ(Nov.ȱ 8,ȱ2017),ȱhttps://www.nytimes.com/2017/11/08/technology/chinaȬfacebook.html.ȱ Wideningȱ theȱ apertureȱ further,ȱ malefactorsȱ atȱ homeȱ andȱ abroadȱ canȱ manipulateȱ socialȱ mediaȱ toȱ promoteȱ extremism.ȱ “Behindȱ everyȱ Facebookȱ ad,ȱ Twitterȱ feed,ȱ andȱ YouTubeȱ recommendationȱ isȱ anȱ algorithmȱ that’sȱ designedȱ toȱ keepȱusersȱusing:ȱItȱtracksȱpreferencesȱthroughȱclicksȱandȱhovers,ȱthenȱspitsȱoutȱaȱ steadyȱstreamȱofȱcontentȱthat’sȱinȱlineȱwithȱyourȱtastes.”ȱKatherineȱJ.ȱWu,ȱRadicalȱ 29ȱ ȱ ȱ IdeasȱSpreadȱThroughȱSocialȱMedia.ȱAreȱtheȱAlgorithmsȱtoȱBlame?,ȱPBSȱ(Mar.ȱ28,ȱ2019),ȱ https://www.pbs.org/wgbh/nova/article/radicalȬideasȬsocialȬmediaȬalgorithms.ȱ Allȱtooȱoften,ȱhowever,ȱtheȱcodeȱitselfȱturnsȱthoseȱtastesȱsour.ȱForȱexample,ȱoneȱ studyȱsuggestsȱthatȱmanipulationȱofȱFacebook’sȱnewsȱfeedȱinfluencesȱtheȱmoodȱofȱ itsȱ users:ȱ placeȱ moreȱ positiveȱ postsȱ onȱ theȱ feedȱ andȱ usersȱgetȱ happier;ȱ focusȱ onȱ negativeȱ informationȱ insteadȱ andȱ usersȱ getȱ angrier.ȱ Adamȱ D.ȱ I.ȱ Kramerȱ etȱ al.,ȱ ExperimentalȱEvidenceȱofȱMassiveȬScaleȱEmotionalȱContagionȱThroughȱSocialȱNetworks,ȱ 111ȱPNASȱ8788,ȱ8789ȱ(2014).ȱThisȱcanȱbecomeȱaȱproblem,ȱasȱFacebook’sȱalgorithmȱ “tendsȱtoȱpromoteȱtheȱmostȱprovocativeȱcontent”ȱonȱtheȱsite.ȱMaxȱFisher,ȱInsideȱ Facebook’sȱ Secretȱ Rulebookȱ forȱ Globalȱ Politicalȱ Speech,ȱ N.Y.ȱ Timesȱ (Dec.ȱ 27,ȱ 2018),ȱ http://www.nytimes.com/2018/12/27/world/facebookȬmoderators.html.ȱ Indeed,ȱ “[t]heȱFacebookȱNewsȱFeedȱenvironmentȱbringsȱtogether,ȱinȱoneȱplace,ȱmanyȱofȱ theȱ influencesȱ thatȱ haveȱ beenȱ shownȱ toȱ driveȱ psychologicalȱ aspectsȱ ofȱ polarization.”ȱJaimeȱE.ȱSettle,ȱFrenemies:ȱHowȱSocialȱMediaȱPolarizesȱAmericaȱ(2018).ȱ Likewise,ȱ YouTube’sȱ videoȱ recommendationȱ algorithm—whichȱ leadsȱ toȱ moreȱ thanȱ 70ȱ percentȱ ofȱ timeȱ peopleȱ spendȱ onȱ theȱ platform—hasȱ beenȱ criticizedȱ forȱ shuntingȱvisitorsȱtowardȱeverȱmoreȱextremeȱandȱdivisiveȱvideos.ȱRooseȱ&ȱConger,ȱ 30ȱ ȱ ȱ supra;ȱseeȱ JackȱNicas,ȱHowȱYouTubeȱDrivesȱPeopleȱtoȱtheȱInternet’sȱDarkestȱCorners,ȱ Wallȱ St.ȱ J.ȱ (Feb.ȱ 7,ȱ 2018),ȱ https://www.wsj.com/articles/howȬyoutubeȬdrivesȬ viewersȬtoȬtheȬinternetsȬdarkestȬcornersȬ1518020478.ȱYouTubeȱhasȱfineȬtunedȱitsȱ algorithmȱ toȱ recommendȱ videosȱ thatȱ recalibrateȱ users’ȱ existingȱ areasȱ ofȱ interestȱ andȱsteadilyȱsteerȱthemȱtowardȱnewȱones—aȱmodusȱoperandiȱthatȱhasȱreportedlyȱ provenȱaȱrealȱboonȱforȱfarȬrightȱextremistȱcontent.ȱSeeȱKevinȱRoose,ȱTheȱMakingȱofȱ aȱYouTubeȱRadical,ȱN.Y.ȱTimesȱ(Juneȱ8,ȱ2019),ȱhttp://www.nytimes.com/interactive/ȱ 2019/06/08/technology/youtubeȬradical.html.ȱȱ ȱ Thereȱ isȱ alsoȱ growingȱ attentionȱ toȱ whetherȱ socialȱ mediaȱ hasȱ playedȱ aȱ significantȱ roleȱ inȱ increasingȱ nationwideȱ politicalȱ polarization.ȱ Seeȱ Andrewȱ Soergel,ȱIsȱSocialȱMediaȱtoȱBlameȱforȱPoliticalȱPolarizationȱinȱAmerica?,ȱU.S.ȱNewsȱ&ȱ WorldȱRep.ȱ(Mar.ȱ20,ȱ2017),ȱhttps://www.usnews.com/news/articles/2017Ȭ03Ȭ20/isȬ socialȬmediaȬtoȬblameȬforȬpoliticalȬpolarizationȬinȬamerica.ȱ Theȱ concernȱ isȱ thatȱ “webȱ surfersȱ areȱ beingȱ nudgedȱ inȱ theȱ directionȱ ofȱ politicalȱ orȱ unscientificȱ propaganda,ȱabusiveȱcontent,ȱandȱconspiracyȱtheories.”ȱWu,ȱRadicalȱIdeas,ȱsupra.ȱ Byȱ surfacingȱ ideasȱ thatȱ wereȱ previouslyȱ deemedȱ tooȱ radicalȱ toȱ takeȱ seriously,ȱ socialȱmediaȱmainstreamsȱthem,ȱwhichȱstudiesȱshowȱmakesȱpeopleȱ“muchȱmoreȱ 31ȱ ȱ ȱ open”ȱtoȱthoseȱconcepts.ȱMaxȱFisherȱ&ȱAmandaȱTaub,ȱHowȱEverydayȱSocialȱMediaȱ Usersȱ Becomeȱ RealȬWorldȱ Extremists,ȱ N.Y.ȱ Timesȱ (Apr.ȱ 25,ȱ 2018),ȱ http://www.nytimes.com/2018/04/25/world/asia/facebookȬextremism.html.ȱ Atȱ itsȱ worst,ȱthereȱisȱevidenceȱthatȱsocialȱmediaȱmayȱevenȱbeȱusedȱtoȱpushȱpeopleȱtowardȱ violence.7ȱTheȱsitesȱareȱnotȱentirelyȱtoȱblame,ȱofȱcourse—theyȱwouldȱnotȱhaveȱsuchȱ successȱ withoutȱ humansȱ willingȱ toȱ generateȱ andȱ toȱ viewȱ extremeȱ content.ȱ Providersȱareȱalsoȱtweakingȱtheȱalgorithmsȱtoȱreduceȱtheirȱpullȱtowardȱhateȱspeechȱ andȱotherȱinflammatoryȱmaterial.ȱSeeȱIsaac,ȱsupra;ȱRooseȱ&ȱConger,ȱsupra.ȱYetȱtheȱ dangersȱofȱsocialȱmedia,ȱinȱitsȱcurrentȱform,ȱareȱpalpable.ȱ Whileȱ theȱ majorityȱ andȱ Iȱ disagreeȱ aboutȱ whetherȱ §ȱ 230ȱ immunizesȱ interactiveȱcomputerȱservicesȱfromȱliabilityȱforȱallȱtheseȱactivitiesȱorȱonlyȱsome,ȱitȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 7ȱSee,ȱe.g.,ȱSarahȱMarsh,ȱSocialȱMediaȱRelatedȱtoȱViolenceȱbyȱYoungȱPeople,ȱSayȱExperts,ȱ Theȱ Guardianȱ (Apr.ȱ 2,ȱ 2018),ȱ https://www.theguardian.com/media/2018/apr/02/socialȬ mediaȬviolenceȬyoungȬpeopleȬgangsȬsayȬexperts;ȱKevinȱRoose,ȱAȱMassȱMurderȱof,ȱandȱfor,ȱ theȱ Internet,ȱ N.Y.ȱ Timesȱ (Mar.ȱ 15,ȱ 2019),ȱ https://www.nytimes.com/2019/03/15/ȱ technology/facebookȬyoutubeȬchristchurchȬshooting.html;ȱCraigȱTimbergȱetȱal.,ȱTheȱNewȱ Zealandȱ Shootingȱ Showsȱ Howȱ TouTubeȱ andȱ Facebookȱ Spreadȱ Hateȱ andȱ Violentȱ Images—Yetȱ Again,ȱ Wash.ȱ Postȱ (Mar.ȱ 15,ȱ 2019),ȱ https://www.washingtonpost.com/technology/ȱ 2019/03/15/facebookȬyoutubeȬtwitterȬamplifiedȬvideoȬchristchurchȬmosqueȬshooting;ȱ Julieȱ Turkewitzȱ &ȱ Kevinȱ Roose,ȱ Whoȱ Isȱ Robertȱ Bowers,ȱ theȱ Suspectȱ inȱ theȱ Pittsburghȱ Synagogueȱ Shooting?,ȱ N.Y.ȱ Timesȱ (Oct.ȱ 27,ȱ 2018),ȱ https://www.nytimes.com/2018/ȱ ȱ10/27/us/robertȬbowersȬpittsburghȬsynagogueȬshooter.html.ȱ 32ȱ ȱ isȱpellucidȱthatȱCongressȱdidȱnotȱhaveȱanyȱofȱthemȱinȱmindȱwhenȱitȱenactedȱtheȱ CDA.ȱTheȱtextȱandȱlegislativeȱhistoryȱofȱtheȱstatuteȱshoutȱtoȱtheȱraftersȱCongress’sȱ focusȱ onȱ reducingȱ children’sȱ accessȱ toȱ adultȱ material.ȱ Congressȱ couldȱ notȱ haveȱ anticipatedȱtheȱperniciousȱspreadȱofȱhateȱandȱviolenceȱthatȱtheȱriseȱofȱsocialȱmediaȱ likelyȱhasȱsinceȱfomented.ȱNorȱcouldȱCongressȱhaveȱdivinedȱtheȱroleȱthatȱsocialȱ mediaȱprovidersȱthemselvesȱwouldȱplayȱinȱthisȱtale.ȱMountingȱevidenceȱsuggestsȱ thatȱprovidersȱdesignedȱtheirȱalgorithmsȱtoȱdriveȱusersȱtowardȱcontentȱandȱpeopleȱ theȱusersȱagreedȱwith—andȱthatȱtheyȱhaveȱdoneȱitȱtooȱwell,ȱnudgingȱsusceptibleȱ soulsȱeverȱfurtherȱdownȱdarkȱpaths.ȱByȱcontrast,ȱwhenȱtheȱCDAȱbecameȱlaw,ȱtheȱ closestȱextantȱancestorȱtoȱFacebookȱ(andȱitȱwasȱstillȱseveralȱbranchesȱlowerȱonȱtheȱ evolutionaryȱtree)ȱwasȱtheȱchatroomȱorȱmessageȱforum,ȱwhichȱactedȱasȱaȱdigitalȱ bulletinȱboardȱandȱdidȱnothingȱproactiveȱtoȱforgeȱoffȬsiteȱconnections.8ȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ ȱ 8ȱSeeȱCaitlinȱDewey,ȱAȱCompleteȱHistoryȱofȱtheȱRiseȱandȱFall—andȱReincarnation!—ofȱ theȱBelovedȱ‘90sȱChatroom,ȱWash.ȱPostȱ(Oct.ȱ30,ȱ2014),ȱhttp://www.washingtonpost.com/ȱ news/theȬintersect/wp/2014/10/30/aȬcompleteȬhistoryȬofȬtheȬriseȬandȬfallȬandȬ reincarnationȬofȬtheȬbelovedȬ90sȬchatroom;ȱ seeȱ alsoȱ Thenȱ andȱ Now:ȱ Aȱ Historyȱ ofȱ Socialȱ NetworkingȱSites,ȱCBSȱNews,ȱhttp://www.cbsnews.com/pictures/thenȬandȬnowȬaȬhistoryȬ ofȬsocialȬnetworkingȬsitesȱ (lastȱ accessedȱ Julyȱ 9,ȱ 2019)ȱ (detailingȱ theȱ evolutionȱ ofȱ socialȱ mediaȱsitesȱfromȱClassmates,ȱlaunchedȱonlyȱ“asȱaȱlistȱofȱschoolȱaffiliations”ȱinȱDecemberȱ 1995;ȱtoȱ“theȱveryȱfirstȱsocialȱnetworkingȱsite”ȱSixȱDegrees,ȱwhichȱlaunchedȱinȱMayȱ1997ȱ ȱ butȱwhoseȱnetworksȱwereȱlimitedȱ“dueȱtoȱtheȱlackȱofȱpeopleȱconnectedȱtoȱtheȱInternet”;ȱ 33ȱ ȱ Whether,ȱ andȱ toȱ whatȱ extent,ȱ Congressȱ shouldȱ allowȱ liabilityȱ forȱ techȱ companiesȱthatȱencourageȱterrorism,ȱpropaganda,ȱandȱextremismȱisȱaȱquestionȱforȱ legislators,ȱnotȱjudges.ȱOverȱtheȱpastȱtwoȱdecadesȱ“theȱInternetȱhasȱoutgrownȱitsȱ swaddlingȱ clothes,”ȱ Roommates.Com,ȱ 521ȱ F.3dȱ atȱ 1175ȱ n.39,ȱ andȱ itȱ isȱ fairȱ toȱ askȱ whetherȱtheȱrulesȱthatȱgovernedȱitsȱinfancyȱshouldȱstillȱoverseeȱitsȱadulthood.ȱItȱisȱ undeniableȱ thatȱ theȱ Internetȱ andȱ socialȱ mediaȱ haveȱ hadȱ manyȱ positiveȱ effectsȱ worthȱ preservingȱ andȱ promoting,ȱ suchȱ asȱ facilitatingȱ openȱ communication,ȱ dialogue,ȱandȱeducation.ȱAtȱtheȱsameȱtime,ȱasȱoutlinedȱabove,ȱsocialȱmediaȱcanȱbeȱ manipulatedȱ byȱ evildoersȱ whoȱ poseȱ realȱ threatsȱ toȱ ourȱ democraticȱ society.ȱ Aȱ healthyȱ debateȱ hasȱ begunȱ bothȱ inȱ theȱ legalȱ academy9ȱ andȱ inȱ theȱ policyȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ toȱFriendster,ȱlaunchedȱinȱMarchȱ2002ȱandȱ“creditedȱasȱgivingȱbirthȱtoȱtheȱmodernȱsocialȱ mediaȱmovement”;ȱtoȱFacebook,ȱwhichȱwasȱ“rolledȱoutȱtoȱtheȱpublicȱinȱSeptemberȱ2006”).ȱ ȱ 9ȱSee,ȱe.g.,ȱDanielleȱKeatsȱCitronȱ&ȱBenjaminȱWittes,ȱTheȱProblemȱIsnȇtȱJustȱBackpage:ȱ Revisingȱ Sectionȱ 230ȱ Immunity,ȱ 2ȱ Geo.ȱ L.ȱ Tech.ȱ Rev.ȱ 453,ȱ 454Ȭ55ȱ (2018);ȱ Jeffȱ Kosseff,ȱ DefendingȱSectionȱ230:ȱTheȱValueȱofȱIntermediaryȱImmunity,ȱ15ȱJ.ȱTech.ȱL.ȱ&ȱPolȇyȱ123,ȱ124ȱ (2010);ȱDanielaȱC.ȱManzi,ȱManagingȱtheȱMisinformationȱMarketplace:ȱTheȱFirstȱAmendmentȱ andȱ theȱ Fightȱ AgainstȱFakeȱ News,ȱ87ȱ Fordhamȱ L.ȱ Rev.ȱ 2623,ȱ 2642Ȭ43ȱ (2019).ȱ Muchȱ ofȱ theȱ enterprisingȱlegalȱscholarshipȱdebatingȱtheȱ intersectionȱ ofȱsocialȱmedia,ȱterrorism,ȱ andȱ theȱCDAȱcomesȱfromȱstudentȱNotes.ȱSee,ȱe.g.,ȱJaimeȱE.ȱFreilich,ȱNote,ȱSectionȱ230’sȱLiabilityȱ ShieldȱinȱtheȱAgeȱofȱOnlineȱTerroristȱRecruitment,ȱ83ȱBrook.ȱL.ȱRev.ȱ675,ȱ690Ȭ91ȱ(2018);ȱAnnaȱ Elisabethȱ Jayneȱ Goodman,ȱ Noteȱ andȱ Comment,ȱ Whenȱ Youȱ Giveȱ aȱ Terroristȱ aȱ Twitter:ȱ HoldingȱSocialȱMediaȱCompaniesȱLiableȱforȱtheirȱSupportȱofȱTerrorism,ȱ46ȱPepp.ȱL.ȱRev.ȱ147,ȱ ȱ 182Ȭ86ȱ (2018);ȱ Nicoleȱ Phe,ȱ Note,ȱ Socialȱ Mediaȱ Terror:ȱ Reevaluatingȱ Intermediaryȱ Liabilityȱ 34ȱ ȱ community10ȱ aboutȱ changingȱ theȱ scopeȱ ofȱ §ȱ 230.ȱ Perhapsȱ Congressȱ willȱ clarifyȱ whatȱIȱbelieveȱtheȱtextȱofȱtheȱprovisionȱalreadyȱstates:ȱthatȱtheȱcreationȱofȱsocialȱ networksȱ reachesȱ beyondȱ theȱ publishingȱ functionsȱ thatȱ §ȱ 230ȱ protects.ȱ Perhapsȱ CongressȱwillȱengageȱinȱaȱbroaderȱrethinkingȱofȱtheȱscopeȱofȱCDAȱimmunity.ȱOrȱ perhapsȱCongressȱwillȱdecideȱthatȱtheȱcurrentȱregimeȱbestȱbalancesȱtheȱinterestsȱ involved.ȱ Inȱ theȱ meantime,ȱ however,ȱ Iȱ cannotȱ joinȱ myȱ colleagues’ȱ decisionȱ toȱ immunizeȱ Facebook’sȱ friendȬȱ andȱ contentȬsuggestionȱ algorithmsȱ fromȱ judicialȱ scrutiny.ȱIȱthereforeȱmustȱinȱpartȱrespectfullyȱdissent,ȱasȱIȱconcurȱinȱpart.ȱ ȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱȱ UnderȱtheȱCommunicationsȱDecencyȱAct,ȱ51ȱSuffolkȱU.ȱL.ȱRev.ȱ99,ȱ126Ȭ30ȱ(2018).ȱ ȱ 10ȱSee,ȱe.g.,ȱTarletonȱGillespie,ȱHowȱSocialȱNetowrksȱSetȱtheȱLimitsȱofȱWhatȱWeȱCanȱSayȱ Online,ȱWiredȱ(Juneȱ26,ȱ2018),ȱhttp://www.wired.com/story/howȬsocialȬnetworksȬsetȬtheȬ limitsȬofȬwhatȬweȬcanȬsayȬonline;ȱ Christianoȱ Lima,ȱ Howȱ aȱ Wideningȱ Politicalȱ Riftȱ Overȱ Onlineȱ Liabilityȱ Isȱ Splittingȱ Washington,ȱ Politicoȱ (Julyȱ 9,ȱ 2019),ȱ http://www.politico.com/ȱ story/2019/07/09/onlineȬindustryȬimmunityȬsectionȬ230Ȭ1552241;ȱ Markȱ Sullivan,ȱ Theȱ 1996ȱ Lawȱ Thatȱ Madeȱ theȱ Webȱ Isȱ inȱ theȱ Crosshairs,ȱ Fastȱ Co.ȱ (Nov.ȱ 29,ȱ 2018),ȱ http://www.fastcompany.com/90273352/maybeȬitsȬtimeȬtoȬtakeȬawayȬtheȬoutdatedȬ loopholeȬthatȬbigȬtechȬexploits;ȱ cf.ȱ Darrellȱ M.ȱ Westȱ &ȱ Johnȱ R.ȱ Allen,ȱ Howȱ Artificialȱ Intelligenceȱ Isȱ Transformingȱ theȱ World,ȱ Brookingsȱ (Apr.ȱ 24,ȱ 2018),ȱ http://www.brookings.edu/research/howȬartificialȬintelligenceȬisȬtransformingȬtheȬ worldȱ(“TheȱmalevolentȱuseȱofȱAIȱexposesȱindividualsȱandȱorganizationsȱtoȱunnecessaryȱ ȱrisksȱandȱunderminesȱtheȱvirtuesȱofȱtheȱemergingȱtechnology.”).ȱ 35ȱ