Quantum physicists recognize that the very act of observing causes an inevitable “disturbance of the object observed.”1 The Supreme Court can sometimes resemble a physicist observing particles when it grants review of a hot legal question: even its decision to say nothing can create ripples in the law. Consider the Court’s decision to place two companion suits against Google and Twitter under its judicial microscope.2 The certiorari grant brewed much anticipation that the Court would, for the first time, shed light on § 230 of the Communications Decency Act of 19963 (CDA).4 But, after almost eighty amicus briefs and three hours of oral argument,5 the Court anticlimactically ducked the § 230 question by disposing of both appeals using the Twitter case, where § 230 had never been litigated.6 Far from a commendable embodiment of judicial minimalism, the Court’s silence on the controversial scope of § 230 made light of the judicial duty “to say what the law is”7 and amplified uncertainties about internet service providers’ liability. By entertaining a critical question only to hold its tongue, the Court engaged in unhelpful judicial ventriloquism — speaking even as it appeared not to — that cast a pall over the apparent victory for social media platforms.
In the first few hours of the year 2017, a gunman named Abdulkadir Masharipov entered the Reina nightclub in Istanbul, Turkey, and “fired over 120 rounds into a crowd of more than 700 people.”8 He “killed 39 people and injured 69 others.”9 An affiliate and trainee of the Islamic State of Iraq and Syria (ISIS), Masharipov had traveled to Turkey in 2016 under the terrorist organization’s orders to launch the attack.10 Among the dead was Nawras Alassaf, a Jordanian citizen who had visited Istanbul with his wife to celebrate the New Year.11
Alassaf’s relatives sued Google, Twitter, and Facebook in the District Court for the Northern District of California.12 They claimed, among other things, that the social media companies were directly liable under the Antiterrorism Act of 200113 (ATA) and indirectly liable under the Justice Against Sponsors of Terrorism Act14 (JASTA).15 The ATA makes it a crime to provide material support to a foreign terrorist organization;16 it further provides for civil liability by allowing a United States national to recover treble damages when injured “by reason of an act of international terrorism.”17 Meanwhile, JASTA attaches indirect liability to any person who “aids and abets” or “conspires with the person who committed such an act of international terrorism.”18
In Taamneh v. Twitter, Inc.,19 the district court granted the defendants’ motion to dismiss for failure to state a claim.20 First, Judge Chen found that the plaintiffs’ direct liability claim failed to adequately allege proximate causation.21 He explained that the ATA’s “by reason of” language required a showing of “some direct relationship” between the plaintiffs’ injuries and the defendants’ acts.22 Conclusory allegations that ISIS’s use of social media radicalized Masharipov fell short of the requirement.23 Turning to indirect liability, Judge Chen concluded that the complaint failed to show the defendants were generally aware of their role in ISIS’s terrorist activities or that they offered substantial assistance.24 The plaintiffs appealed the dismissal of their JASTA claim.25
The Ninth Circuit reversed.26 Writing for the panel, Judge Christen27 concluded that the plaintiffs’ complaint satisfied the tripartite test for aiding and abetting liability under JASTA.28 She reviewed the three elements outlined in Halberstam v. Welch,29 which JASTA specified as “the proper legal framework”30: (1) the aided principal must “perform a wrongful act” causing injury, (2) the defendant must be “generally aware” of his or her role in the illegal activity, and (3) the defendant must “knowingly and substantially assist the principal violation.”31 The parties did not dispute that the first element was satisfied.32 Next, persistent media coverage of and governmental pressure concerning ISIS’s social media use established the defendants’ general awareness.33 The plaintiffs also satisfied the knowing-assistance prong of the third element by alleging that the defendants “refused to take meaningful steps” to prevent ISIS from promoting and facilitating terrorist activities by using their social media platforms.34 As for the substantial-assistance prong, the court went through the six factors outlined in Halberstam to conclude that the plaintiffs cleared this requirement as well by alleging “that defendants provided services that were central to ISIS’s growth and expansion . . . over many years.”35
In the same opinion, the court also addressed an appeal from Gonzalez v. Google, Inc.,36 which concerned a similar aiding and abetting claim against Google for the 2015 ISIS attack in Paris.37 The district court held that § 230 of the CDA barred most of the plaintiffs’ claims.38 Section 230 provides, in part, that “[n]o provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.”39 The Ninth Circuit affirmed.40 Judge Christen rejected the plaintiffs’ theory that Google’s algorithmic recommendation of ISIS-related content amounted to content created by Google.41 Citing precedents, she explained that such content-neutral algorithms could not expose an internet service provider to liability “for content posted by a third-party.”42
The Supreme Court reversed the Ninth Circuit’s holding in the Twitter case.43 Writing for a unanimous court, Justice Thomas began by clarifying the relevant legal framework for determining “whether defendants’ conduct constitute[d] ‘aid[ing] and abett[ing], by knowingly providing substantial assistance.’”44 First, he defined what it means to aid and abet.45 Invoking the presumption that familiar common law concepts “brin[g] the old soil” with them,46 the Court surveyed the development of aiding and abetting doctrine in criminal and tort law.47 Criminal law’s concern about “boundless”48 liability and its emphasis on “truly culpable conduct”49 informed its counterpart in tort law.50 Because Halberstam “reflected and distilled those common-law principles,”51 the Court reasoned, the phrase “aids and abets” in JASTA had to “refer to a conscious, voluntary, and culpable participation in another’s wrongdoing.”52 Second, Justice Thomas addressed the question of “what precisely a defendant must aid and abet.”53 Brushing aside the parties’ dispute about the syntax of the statutory text, he stressed that “aiding and abetting is inherently a rule of secondary liability for specific wrongful acts.”54 As such, defendants must abet “the act of international terrorism that injured the plaintiffs,” not the general enterprise of ISIS.55
The Court concluded that the plaintiffs’ aiding and abetting claim failed on the third Halberstam element.56 Allegations that the defendants had created platforms available to ISIS and used algorithms that displayed content based on user inputs and user history amounted to “mere passive nonfeasance.”57 The Court stressed that the defendants had not given ISIS any special treatment; the platforms were “generally available to the internet-using public”58 and the algorithms were “agnostic as to the nature of the content.”59 The complaint also showed no reason to believe that the companies had consciously participated in the attack.60 Although the Court left open the possibility that a different set of facts could justify holding a social media company liable,61 it refused to adopt “the expansive scope of plaintiffs’ claims,”62 which could attach indirect liability to all internet or cell service providers.63
The Court faulted the Ninth Circuit for misapplying the Halberstam framework.64 The Ninth Circuit erred in (1) focusing on the defendants’ assistance to ISIS generally, rather than whether defendants had aided the Reina attack; (2) conflating the defendants’ general awareness of ISIS’s presence on their platforms with whether they had knowingly assisted terrorism; and (3) treating the six factors on substantial assistance as “disparate, unrelated considerations” rather than guidelines to “capture the essence of aiding and abetting,” which is “participation in another’s wrongdoing that is both significant and culpable enough to justify attributing the principal wrongdoing to the aider and abettor.”65
Justice Jackson briefly concurred.66 Writing alone, she noted the Court’s narrow holding was limited to the cases’ particular circumstances.67 She also clarified that the common law principles invoked to illuminate JASTA “[did] not necessarily translate to other contexts.”68
On the same day it decided Twitter, Inc. v. Taamneh,69 the Court released a per curiam opinion in Gonzalez v. Google LLC.70 Because the indirect liability claims were “materially identical to those at issue in Twitter,” the Court explained, “plaintiffs’ complaint — independent of § 230 — state[d] little if any claim for relief.”71 The Court expressly declined to address the question of whether § 230 barred the indirect liability claims.72
Described as the “twenty-six words that created the Internet,”73 47 U.S.C § 230(c)(1), which immunizes internet service providers from civil liability for content that they host, has long been criticized by both sides of the political aisle. The left blames it for rampant misinformation and extremist speech; the right, for censorship of conservative viewpoints.74 Twitter was the Court’s way of deflecting that hot potato: instead of using Gonzalez as the vehicle for clarifying § 230, the Court elected to resolve both cases via Twitter, where § 230 had never been raised in the lower courts. While some may laud Gonzalez and Twitter as a healthy exercise of judicial minimalism,75 the Court’s silence on the immunity issue exemplifies judicial irresponsibility that will likely fan anxieties about platform liability.
Judicial minimalism is the principle that judges should “say no more than necessary to justify an outcome.”76 Professor Cass Sunstein, often credited with developing the contemporary version of the principle,77 expounds it along two axes: breadth and depth.78 Minimalist judges should decide a case both narrowly and shallowly — meaning, respectively, that they should address only the facts of the case without establishing sweeping rules and should reach a concrete agreement without theorizing about fundamental principles.79 Proponents tout minimalism for reasons of judicial prudence, institutional capacity, and democratic legitimacy. Narrow decisions reduce the costs of adjudication and mistaken judgments, giving “the democratic process room to adapt to future developments.”80 Shallow decisions allow political actors to resolve the deeper disputes.81
Upon first glance, the pair of decisions in Twitter and Gonzalez may fit the bill. Twitter seems narrow: Justice Thomas emphasized that the Court need not resolve whether a different fact pattern involving a clearer duty to monitor or more affirmative assistance from platforms might trigger aiding and abetting liability.82 Justice Jackson stressed the narrowness of the decision even more clearly.83 And Gonzalez looks shallow: the Court avoided the (deep) theoretical question of whether Google should win only because the allegations failed to support liability or also because platforms enjoy immunity against civil claims.84
The bare-bones rulings in Twitter/Gonzalez may seem particularly prudent in their caution in exploring the digital landscape of § 230 immunity, which does not extend to content that an internet service provider itself creates.85 Deciding whether algorithmic recommendations lose § 230 protection may require a technical understanding of algorithms. As Justice Kagan frankly admitted, the Justices are “not like the nine greatest experts on the Internet.”86 Under this view, then, Twitter/Gonzalez was an exemplar of judicial minimalism, simultaneously avoiding the risk of erroneous judgment on a technical question with far-reaching consequences and leaving the politically contentious issue of § 230’s scope to the democratically accountable Congress.
Using judicial minimalism to compliment a decision, however, is a dangerous habit for at least two reasons. First, as a descriptive matter, “minimalism” risks being a hollow label, because most articulations of it fail to specify the metric along which a judgment should be minimal.87 For example, what if a minimalist avoidance of a question unnecessarily deviates from the settled decisional order of a legal inquiry? When facing a legal issue with multiple questions, courts often converge on a logical timeline to answer one question first consistently, especially when that order accords with legislative intent. Platform liability is a case in point. As a matter of logic, the need to judge whether a plaintiff has stated a claim is contingent on answering first whether § 230 protects the platform being sued at all. As for legislative intent, the whole point of § 230 is to shield platforms from even the threat of civil liability so as to avoid chilling free speech.88 Section 230 immunity is therefore “generally accorded effect at the first logical point in the litigation process.”89 Lower courts have consistently treated § 230 immunity as a gateway inquiry before examining whether the theory of liability holds up.90 For example, in Klayman v. Zuckerberg,91 the court dismissed a negligence suit against Facebook solely on § 230 grounds and omitted any discussion of the elements of liability,92 despite the fact that the defendants had brought up the liability argument in their motion to dismiss.93
Twitter/Gonzalez disrupted the consensus, creating doubts about how far § 230 stretches and raising decision costs for courts as they reassess the correct order of procedure. Now, lower courts must weigh just how uncertain they need to be about the anterior question of immunity to skip it and dispose of the case on posterior questions of liability.
On that metric of decisional order, then, it is much less clear that Twitter/Gonzalez was minimalist. The Supreme Court could have hewed close to the accepted order by at least addressing § 230. Given that the plaintiffs’ theory hinged on the narrow argument that a YouTube thumbnail includes information provided by the website and thus loses § 230 protection for third-party content,94 the Court could have issued a narrow opinion without establishing sweeping rules about all forms of algorithmic recommendations. Instead, the Court departed from the norm, implicitly abrogating the spirit of § 230 by indicating that immunity need not be treated as the bottleneck that controls theories of liability. Twitter/Gonzalez may not have said this explicitly, but it may have signaled “more than necessary to justify an outcome.”95
One might nonetheless consider it minimalist for the Court to rule on the cases via the “easier” route of exposing the far-fetched nature of the plaintiffs’ allegations.96 Assuming arguendo that adjudicating the liability question did incur lower decision costs,97 such thinking glosses over the core purpose of § 230. Immunity and liability in § 230 are not conjunctive factors of a test; rather, the immunity analysis was meant to preclude the liability question. If the Court’s deviation from that decisional order is at least plausibly considered an invitation for lower courts to ponder the wisdom of § 230, then Twitter/Gonzalez raises decision costs — just what minimalism is not supposed to do.
Setting aside the descriptive problem of whether Twitter/Gonzalez can properly be labeled minimalist, there is also a jurisprudential problem. Minimalism stands fundamentally in tension with the Supreme Court’s duty to guide lower courts.98 Where there is great uncertainty as to how a law should apply — caused by, for example, new technologies — the Court’s silence merely exports the costs of uncertainty to lower courts and litigators who must “resolve the unanswered questions later.”99 Section 230 immunity presents just such circumstances.
One looming question, for example, is whether algorithmic recommendations receive § 230 protection. In another case involving an ATA claim against Facebook, the Second Circuit had answered yes, following most circuits in interpreting § 230 as granting broad immunity.100 By contrast, the “definitional” interpretation of § 230, which treats subsection (c)(1) as merely defining “publisher,” may well dictate the opposite answer. This approach was adopted by the Seventh Circuit101 and endorsed by Justice Thomas.102 Given the ubiquity of algorithms and their importance to the operation of many websites, less guidance on the question means that lower courts and social media platforms will continue to wrestle with uncertainties about legal liability. And these uncertainties matter: just the specter of legal battles can amount to “death by ten thousand duck-bites,”103 especially for platforms that lack the formidable legal teams of Google and Twitter.
Indeed, the Court’s silence on the scope of immunity may cause stress among internet service providers as they try to figure out what the Court is trying to say by not saying. When the Court grants certiorari, it does so presumably because it recognizes the need to clarify a legal question.104 As such, using Twitter to deflect the central question of Gonzalez can be interpreted as a signal that § 230 immunity is often unnecessary because the liability issue operates as an adequate backstop. The opinion also included what reads like “coaching” language for future plaintiffs: “Try us again with a better fact pattern, and a well-supported theory of civil liability might pierce § 230 immunity!”105
The brunt of this uncertainty will fall on not only smaller platform companies but also marginalized communities, who will face disparately greater censorship by platforms that want to “play it safe.” Because getting sued is expensive while censoring speech is free, the “profit-maximizing intermediary likely will choose the mechanism that is least costly, rather than the one that preserves the most speech.”106 Bias will contaminate monitoring and removing content, with or without the aid of automated processes.107 For example, content concerning and from the Middle East is often removed due to its erroneous association with terrorism.108 Neglecting to address uncertainties with such disparate consequences reveals Twitter’s “prudence” to be judicial irresponsibility.
One might still insist that institutional capacity and democratic legitimacy justify the Court’s avoidance of the § 230 question. Not so. First, while the Justices may not have PhDs in computer science, neither are they totally at sea when contemplating technology. Van Buren v. United States109 offers inspiration. There, the Court tackled the technical question of interpreting what it means to have “authorized access” to a computer under the Computer Fraud and Abuse Act of 1986110 (CFAA), relying on traditional tools of textualism.111 Just as the CDA was enacted long before the birth of social media, so too the CFAA long predated sophisticated methods of web authentication.112 Determining whether an automated curation of content falls within § 230’s language is an exercise not in software engineering, but in statutory interpretation. At a minimum, the Justices surely have expertise in settling whether a subsection of a statute is substantive or definitional.113
Second, the argument from democratic legitimacy has bite only if Congress can fix the problem. Unfortunately, efforts to amend § 230 have been fruitless.114 It was up to the Court to clarify the boundaries of a website’s protected functions. It left no breadcrumbs at all.
Many commentators were quick to celebrate Twitter/Gonzalez.115 From the perspective of protecting free speech and preserving the internet, that is an understandable reaction. However, the Court’s cavalier postponement of the § 230 question casts a pall over platform providers, who now have to worry about losing the war even as they won a battle. Couched in the language of judicial self-restraint, Twitter and Gonzalez missed an important opportunity to clarify a pressing area of internet law while sending ominous signals about platform liability. Whatever this decision means, it is not a paradigm of judicial minimalism — at least, not a good one.