Law & Economics Scholars Amicus to US District Court in United States v Google
INTERESTS OF AMICI CURIAE
Amici are eighteen scholars of antitrust law and economics at leading universities and research institutions across the United States. Their names, titles, and academic affiliations are listed in the Appendix. All possess expertise in, and collectively have conducted extensive research on, antitrust law and economics, including the competitive dynamics of digital markets and the appropriate scope of antitrust remedies.
Amici have an interest in ensuring that antitrust law promotes the public interest by remaining grounded in sensible rules informed by sound economic analysis. That includes ensuring that any remedy in this case promotes consumer welfare and competition without unduly punishing success or undermining innovation. We file this brief to assist the Court in evaluating the economic and legal implications of the remedies proposed in this case, and to caution against overbroad measures that exceed the proven harms.[1]
SUMMARY OF ARGUMENT
Plaintiffs chose to pursue liability based on inferences drawn from Google’s search-distribution agreements and their purported effect on user choice and rival scale. Having persuaded the Court to premise liability on that basis, Plaintiffs cannot now escape the consequences of their chosen path, including the need for a stronger showing of causation at the remedy phase.
As the D.C. Circuit has observed, “relief should be tailored to fit the wrong creating the occasion for the remedy.” United States v. Microsoft, 253 F.3d 34, 107 (D.C. Cir. 2001) (“Microsoft”). And, as that court also admonished, a court “must base its relief on some clear ‘indication of a significant causal connection between the conduct enjoined or mandated and the violation found directed toward the remedial goal intended.’” Id. at 105 (quoting 3 Phillip E. Areeda & Herbert Hovenkamp, Antitrust Law, ¶ 653b, at 91-92 (1996)). That caution remains even where, as here, courts infer causation of competitive harm in finding liability. Mem. Op., U.S., et al. v. Google LLC, 20-cv-3010, ECF No. 1033 (“Mem. Op.”) at 216 (citing Microsoft, 253 F.3d at 79). In brief, the Circuit Court’s decision on remedies in Microsoft suggests that there is no “edentulous” standard for remedies in cases brought under Section 2 of the Sherman Act.
Courts should take special care to err on the side of restraint in cases where unlawful conduct may have both pro- and anticompetitive effects, and where the magnitude of competitive harm caused by unlawful conduct is difficult to ascertain. Antitrust remedies must address the specific exclusionary conduct proven, not serve as a vehicle to punish the defendant or to implement the speculative industrial organization designs preferred by regulators.
These general principles are ignored in Plaintiffs’ Revised Proposed Final Judgment (ECF No. 1184-1). Most of Plaintiffs’ wide-ranging proposals are untethered from findings of harm and would neither stop nor remediate the conduct found unlawful in the liability phase of this trial. Indeed, most of the proposed remedies appear designed as penalties, rather than remedies—and suboptimal penalties at that. They risk considerable harm to competition, to the quality of Google’s general search engine (GSE) and other products, and to the more than 200 million U.S. consumers who derive benefits from what is “widely recognized as the best GSE available in the United States.” Mem. Op. at 46 ¶ 126.
Instead, the Court should craft relief narrowly tailored to the specific conduct found to be anticompetitive—for example, by prohibiting truly exclusive deals or coercive tying—while avoiding complex, long-running, behavioral remedies or breakups that do not directly address harm cause by the violative conduct. This measured approach aligns with the approach taken by the D.C. Circuit in Microsoft and adopted by the Supreme Court in Verizon Communications, Inc. v. Law Offices of Curtis Trinko, 540 U.S. 398 (2004). The Microsoft and Trinko decisions both counsel caution in imposing burdensome affirmative obligations or structural changes absent a but-for causal nexus to the violation found and the harm caused by it. That approach is sound. Remedies should target specific anticompetitive acts without deterring the competitive process that benefits consumers.
ARGUMENT
I. Remedies Must Address Harm to Competition Caused by Unlawful Conduct
Antitrust remedies in Section 2 cases are not penalties, and neither precedent nor economics contemplates remedies untethered from the conduct found unlawful or the harm caused by that unlawful conduct.
Plaintiffs’ proposed remedies violate these core principles. For a start, they fail to meet the crucial requirement that antitrust remedies should address only the harm caused by anticompetitive conduct. This is particularly true for Plaintiffs’ call to force the divestiture of the Chrome browser—which was developed prior to the agreements at issue here, and became popular by offering a superior product in competition with other browsers. Chrome was not found to be a monopoly asset, and Google’s ownership and operation of Chrome do not depend on any search distribution agreements. Demanding the divestiture of Chrome would not remedy the exclusive search deals at issue. Instead, it would sacrifice both organizational and production efficiencies that benefit millions of consumers.
Moreover, the Court found that Google’s search distribution agreements were exclusionary to the extent they harmed competition by foreclosing rivals. But there was no proof that, “but for” those deals, a competitor would have achieved competitively meaningful scale. A remedy that aims to ensure the success of competing search engines (rather than their ability to compete) would ignore this distinction.
As the D.C. Circuit observed in Microsoft, “relief should be tailored to fit the wrong creating the occasion for the remedy.” Microsoft, 253 F.3d at 107; see also Herbert J. Hovenkamp, Structural Antitrust Relief Against Digital Platforms, 7 J. Law & Innovation 57, 64 (2024) (“the need to get it right—to avoid both under- and over-deterrence—is one of antitrust’s most vexing problems. It is nowhere more pronounced than in the antitrust law of remedies”). To ignore this principle risks doing more harm than good. “Fashioning appropriate equitable antitrust relief requires that courts balance the benefit to competition against the hardship or competitive disadvantage the remedy may cause.” Ginsburg v. InBev NV/SA, 623 F.3d 1229, 1235 (8th Cir. 2010). And this concern is not merely theoretical: The history of antitrust remedies shows that they fail precisely when they over-index on harms and ignore the benefits that may also arise from ambiguous conduct and complex market structures. See generally Robert W. Crandall & Kenneth G. Elzinga, Injunctive Relief in Sherman Act Monopolization Cases, 21 Res. Law and Econ. 277, 33537 (2004) (studying effects of behavioral remedies imposed in ten major monopolization cases). “Without a firm grasp of the economic forces that are driving changes in market structure, the DOJ cannot be expected to design ‘relief’ that will result in increased competition, lower prices, and consumer benefits.” Id. at 335.
In economic terms, remedies calibrated to the harm done are efficient because they force firms to internalize the harm that their conduct has caused, creating incentives for firms to avoid such harmful conduct going forward. Where the conduct at issue implicates both benefits and harms—as with the default installation and placement of Google Search (Google’s general search engine or “GSE”) and other desirable apps—the chilling of pro-competitive conduct and consumer benefits make both over- and under-inclusive remedies a concern: limitations (or additions) to the scope or magnitude of the remedy are liable to be inefficient. See generally, Steven Shavell, Strict Liability Versus Negligence, 9 J. Legal Stud. 1 (1980). Remedies for such conduct can, in that regard, be contrasted with penalties for conduct that is unequivocally harmful, where courts and enforcers have little or no concern with overdeterrence. Compare Louis Kaplow, The Optimal Probability and Magnitude of Fines for Acts that Definitely are Undesirable, 12 Int. Rev. Law & Econ. 3 (1992) with Robert Cooter, Prices and Sanctions, 84 Colum. L. Rev. 1523, 1529 (1984).
This is particularly relevant here, as the Court’s finding of liability was based in part on extrapolations regarding the “stickiness” of defaults, rooted in general observations about the behavioral psychology of consumers. While the Court found these sufficient to support liability, both the Court’s findings of fact and the underlying research recognize that most consumers prefer Google Search to available alternatives, regardless of the default option provided. See, e.g., Mem. Op. at 46 (“Google is widely recognized as the best GSE available in the United States.”); id. at 229 (“Google is the dominant GSE [on Windows devices], even though Windows devices come preinstalled with Microsoft’s Edge browser, which defaults to Bing.”). Saddling users with inferior default GSEs will impose costs on hundreds of millions of consumers, either by forcing them to incur the cost of switching back to Google or by steering them to bear the costs of using an inferior GSE.
To impose a remedy based on inferred competitive harm—and, further, one rooted in inferred and non-quantified conclusions about consumer behavior—implies a fundamental cautionary challenge: without more, it is difficult for a court to know whether the remedy will yield benefits that exceed its costs. Where findings rely on complex economic theories, involve conduct with potential efficiencies, or leave uncertainty about the scope and magnitude of competitive harm, broad or radical remedies are inappropriate. See Microsoft, 253 F.3d at 78-80 (“[D]ivestiture is a remedy that is imposed only with great caution, in part because its long-term efficacy is rarely certain. Absent some measure of confidence that there has been an actual loss to competition that needs to be restored, wisdom counsels against adopting radical structural relief.”).
It is for these reasons that “[a] court … must base its relief on some clear ‘indication of a significant causal connection between the conduct enjoined or mandated and the violation found directed toward the remedial goal intended.’” Id. at 105 (quoting Areeda & Hovenkamp at ¶ 653b, 91-92). The Microsoft court’s emphasis on the need to consider “whether plaintiffs have established a sufficient causal connection,” id. at 106, between the anticompetitive conduct and the defendant’s dominance is especially telling given the lower bar set for liability in that case—the “edentulous standard” applied given numerous findings of harm to a “nascent competitor.” Hence, on remand in Microsoft the Court of Appeals directed the District Court to “consider whether plaintiffs have established a sufficient causal connection between Microsoft’s anticompetitive conduct and its dominant position in the [operating system] market…Absent such causation, the antitrust defendant’s unlawful behavior should be remedied by ‘an injunction against continuation of that conduct.’” Id. (quoting Areeda & Hovenkamp at ¶ 650a, 67). To put it another way, there is no “edentulous standard” for remedies in a Section 2 case. Restricting Google’s conduct without convincing evidence that it is the true source of consumer harm is akin to prescribing strong medicine for an ailment that hasn’t been firmly diagnosed: you might kill the fever, or you might kill the patient.
By the same token, courts and enforcers alike should beware nirvana fallacies, as visions of ideal competition may be out of reach—or, in fact, relatively impoverished. See, e.g., A. Douglas Melamed, Afterword: The Purposes of Antitrust Remedies, 76 Antitrust L.J. 359, 368 (2009) (“[R]emedies are hard to get right and, when suboptimal, can undermine antitrust objectives by interfering with markets and prohibiting or deterring procompetitive conduct.”). A court’s task is not to engineer an ideal state of GSE competition, because, as the Supreme Court observed in Trinko, central planning is “a role for which [the courts] are ill-suited.” Trinko, 540 U.S. at 408.
For one thing, the conditions that lead to relatively concentrated markets may persist no matter what the intervention. Neither economics nor precedent suggest that an imagined state of atomistic competition should be preferred to an actual competitive market, even if dominated by a few, or even one, firm. More is not always better—not as measured by standard competition-relevant metrics associated with consumer welfare: price, output, innovation, and quality. Antitrust law recognizes that even monopoly can arise “as a consequence of a superior product, business acumen, or historic accident.” United States v. Grinnell Corp., 384 U.S. 563, 571 (1966). For that reason, antitrust law and, specifically, Section 2, do not condemn monopoly itself. Id. at 570-71.
Moreover, network effects, like scale advantages, are not inherently anticompetitive. In fact, direct network effects arise when the value of a product (or platform) increases as the number of its consumers increase. See generally David S. Evans & Richard Schmalensee, Network Effects: March to the Evidence, not to the Slogans, CPI Antitrust Chronicle (Aug. 2017) (reviewing economic research regarding network effects). Network effects, by definition, confer consumer benefits, and well-tailored remedies must account for those benefits.
The failure of the remedy in the General Motors bus case is instructive. Crandall & Elzinga, supra, at 317-35 (discussing U.S. v. General Motors Corp., Consent Decree, C.C.H Trade Cases ¶ 71,624 (1965)). There, “an attempt to force additional competition among multiple firms into such a small market with large economies of scale proved to be futile.” Id. at 323. Despite comprehensive behavioral remedies aimed not only at prohibiting GM’s allegedly exclusionary conduct, but also at facilitating entry by other firms, competition was stymied by the realities of the market’s scale requirements. Instead, “[t]he consent decree may have facilitated the unsustainable entry [of a new competitor].” Id. at 325 (emphasis added). Even the provisions aimed merely at prohibiting past, allegedly exclusionary conduct failed to account for the benefits of scale: “GM was barred by the decree from entering into long-term exclusive contracts with Greyhound, thereby depriving it of the assured benefits of scale necessary for efficient operation. GM therefore exited from intercity bus manufacturing in 1979.” Id. at 324.
Finally, neither network effects nor scale advantages render markets fundamentally uncontestable. In what are sometimes called “reverse network effects,” the same principles that can lead to growth and concentration can have the opposite effect, as “[e]ach lost customer induces other customers to leave, which induces more to leave.” Evans & Schmalensee, supra, at 3; see also Joseph Farrell & Paul Klemperer, Coordination and Lock-In: Competition with Switching Costs and Network Effects, in Vol. 3 Handbook of Industrial Organization (Mark Armstrong & David E.M. Sappington, eds., 2007) (reviewing literature on network effects and competition and noting that “competition for the market” or “life-cycle competition” can replace ordinary compatible competition). Examples abound. “Myspace demonstrates this perfectly: once users began leaving, network effects operated in reverse, hastening its collapse.” Brian C. Albrecht, Network Effects in FTC v. Meta, Truth on the Market (Apr. 16, 2025), available at https://truthonthemarket.com/2025/04/16/network-effects-in-ftc-v-meta/.
II. Structural Remedies to Monopolization Claims Are Rightly Disfavored
The difficulties and risks associated with structural remedies in Section 2 cases have been well documented. See, e.g., Hovenkamp, supra, at 64 (“Unnecessary or badly designed breakups can do a great deal of harm, particularly in highly innovative markets. This is doubly true about the breakup of internally developed as opposed to acquired assets.”). History, meanwhile, shows that such breakups rarely yield consumer gains. See generally Robert W. Crandall, The Failure of Structural Remedies in Sherman Act Monopolization Cases, 80 Or. L. Rev. 109 (2001) (assessing the success or failure of the major Section 2 cases ending in divestiture). Indeed, there is “remarkably little evidence that these cases and the relief that emanated from them had a positive effect on competition and consumer welfare.” Id. at 197. The reasons are revealing: the failure to account for rapidly changing market forces; the faulty design of remedies that failed to generate an increase in price competition; erroneous assumptions about the competitive effects of increasing the number of competitors; and the imposition of vertical divestitures that actually increased prices or reduced competition by eliminating competitors from the market. Id. The consistent theme emerging from this history is the profound difficulty courts face in successfully re-engineering complex industries through antitrust remedies. “By and large, the antitrust record of monopoly breakup decrees has not been pretty. The courts certainly have the power to ruin a firm, but coming up with a structural remedy that will actually increase output or improve the welfare of consumers or labor is not easy.” Hovenkamp, supra, at 57.
Plaintiffs propose that “Google must promptly and fully divest Chrome, along with any assets or services necessary to successfully complete the divestiture, to a buyer approved by the Plaintiffs in their sole discretion, subject to terms that the Court and Plaintiffs approve.” ECF No. 1184-1 at 12. Divestiture is often described as the “most drastic” of antitrust remedies. See, e.g., United States v. E. I. du Pont de Nemours & Co., 366 U.S. 316, 326 (1961). Requiring Google to divest Chrome would be a drastic remedy indeed.
Chrome was not found to be a monopoly asset, either lawful or unlawful, and Google’s mere ownership and operation of Chrome does not depend on any search-distribution agreements: Demanding its divestiture would not directly address the exclusive search deals at issue. As this Court has previously recognized, Supreme Court precedent establishes that “related acts must also be ‘unlawful’ or of the ‘same type or class’ in order to warrant injunction,” and does not permit “clearly lawful practices [to] be enjoined simply because they will weaken the antitrust violator’s competitive position.” New York v. Microsoft Corp., 224 F. Supp. 2d 76, 109 (D.D.C. 2002) (citing Zenith Radio Corp. v. Hazeltine Research, Inc., 395 U.S. 100, 132-33 (1969)).
While divestiture might benefit a single acquiring competitor, it would do so to the detriment of Chrome’s ongoing development. As a result, the longer-term viability of Chrome might be imperiled. Even if some competitors would benefit from the divesture of Chrome, that remedy would provide no clear benefit to consumers, or to search competition more broadly.
Chrome is much more than a stand-alone piece of software: It is deeply intertwined with Google’s development infrastructure, security and anti-malware tools, and revenue model (search ads and related services). The divestiture of Chrome—and the attendant isolation of Chrome and Google Search—would thus impose technical limitations beyond mere implementation details. The integration of Chrome and Google likely generates real efficiencies in development, distribution, operation, and production. These synergies give Google the incentive and ability to invest heavily in Chrome—funding rapid iteration, security patches, speed improvements, and experimental features like privacy sandboxes. It also allows Chrome to remain free for users while leveraging revenue from Google Search deals. A forced separation would not only incur massive direct costs but could also destroy these efficiencies, potentially leading to a less capable or less secure browser for end-users.
Divestiture would also imperil the ongoing viability of competition in the browser market, especially in conjunction with restrictions on Google payments for default status. See Mem. Op. at 116 ¶ 335 (“Google’s 2021 revenue share payment to Mozilla was over $400 million, or about 80% of Mozilla’s operating budget” and “Mozilla has repeatedly made clear that without these payments, it would not be able to function as it does today”); Gregory J. Werden, Harm to the Competitive Process in the Google Search Case, Mercatus Center Antitrust & Competition Working Paper (Jul. 9, 2024), at 28, available at https://www.mercatus.org/research/working-papers/harm-competitive-process-google-case (“Without the revenue from Google, Mozilla could discontinue support for Firefox, especially if Bing does not greatly improve with its immediate boost in scale.”).
It is for precisely these sorts of reasons that structural remedies are disfavored in Section 2 cases like this one. While structural remedies can potentially be justified in merger cases because “clear demarcations between entities and units [may] still exist, facilitating a structural solution,” the cost and risk of imposing structural remedies in long-established firms is much higher. Thomas O. Barnett, Section 2 Remedies: What to Do After Catching the Tiger by the Tail, 76 Antitrust L.J. 31, 39 (2009). “This difference greatly heightens the risk that a Section 2 structural remedy will create market inefficiencies.” Id.
III. Imposing a Data-Sharing ‘Duty to Deal’ Would Exceed Legal Boundaries and Undermine Innovation
The Supreme Court’s decision in Trinko is the lodestar here. In that case, the Court made clear that antitrust law does not, as a general matter, require a monopolist to aid its competitors by sharing infrastructure or data, warning that such forced sharing “may lessen the incentive for the monopolist, the rival, or both to invest” in valuable innovations. Trinko, 540 U.S. at 408.
Yet Sections VI, VII, and VIII of Plaintiffs’ Revised Proposed Final Judgment describe an astonishing array of mandated information sharing involving large and diverse data sets, data structures, algorithms, models, applications, application programming interfaces (APIs), and other intellectual property. Forcing Google to share proprietary search data, tools, and other intellectual property as proposed would impose an extreme—indeed unprecedented—duty to deal that is at odds with modern antitrust jurisprudence since Trinko.
The resources to be shared with or licensed to competitors “at marginal cost” would include, inter alia, Google’s search index, “scale-dependent data inputs” for search and advertising, user-side data, ad data, training data, data structures, statistical models, apps, APIs, and much more. Proposed Final Judgement (ECF No. 1184-1) §§ VI – VIII. Moreover, mandated sharing would not be confined to extant information resources. Google would also be required to facilitate competitors’ access to, and use of, such resources; and Google would be required to share new information resources, including data and technology, not yet available.
Some of Google’s competitive advantage is related to the scale of its data, and the Court found that some of that scale advantage has been achieved or maintained by the search default agreements found exclusionary at trial. Plaintiffs appear to conclude that, therefore, any and all present, would-be, and—for periods of five and ten years—future competitors should have ready access to any of Google’s intellectual property that is somehow related to the size, scope, and quality of Google’s information resources and the size of the firm’s installed base of users. That is not relief “tailored to fit the wrong creating the occasion for the remedy.” Microsoft, 253 F.3d at 107.
This kind of compelled-access remedy (a “data-sharing” or “interoperability” mandate) raises serious practical and legal concerns. Forcing a firm to hand over the “source of its advantage” to rivals risks chilling investment and turning courts into regulators of ongoing business relations. Post-Trinko, courts are extremely reluctant to impose any affirmative duty to deal on a monopolist absent limited circumstances not present here. See, e.g., Metronet Services Corp. v. Qwest Corp., 383 F.3d 1124, 1133 (9th Cir. 2004) (explaining that only where the harm at issue is redressable by a remedy that “simply order[s] the defendant to deal with its competitors on the same terms”—manifestly not the case here—can a court potentially escape the remedial ordeals discussed in Trinko).
Indeed, forcing Google to share its search index, real-time search data, or user signals would be even more invasive than the network-sharing obligations at issue in Trinko. Google’s search algorithms and user data are highly proprietary, and the product of billions of dollars of R&D investment over two decades. Notably, the D.C. Circuit’s decision in Massachusetts v. Microsoft Corp. rejected forced, royalty-free sharing of extensive technical information with rivals, and the open-sourcing of Internet Explorer (among other things), precisely because those proposed remedies would have deprived Microsoft of valuable intellectual property. See Massachusetts v. Microsoft Corp., 373 F.3d 1199, 1230-31 (D.C. Cir. 2004) (holding that a “‘royalty-free, non-exclusive perpetual right’ of others to use [Microsoft’s intellectual property] would confiscate much of the value of Microsoft’s investment” and was an “analogous form of structural relief” to divestiture).
At the same time, Plaintiffs’ proposal that copious resources be provided to rivals “at marginal cost” (ECF No. 1184-1 at 15) seems to suppose—erroneously—that all of Google’s fixed costs are sunk costs, ignoring, e.g., billions of dollars in annual investments in research and development, and, indeed, hardware and infrastructure, by both Google and its competitors. The problem of investment incentives is especially significant in industries with high (and ongoing) fixed costs that could not be recouped under the proposed interoperability and sharing requirements. As the Court itself noted, “[b]uilding and maintaining a competitive GSE require an extraordinary upfront capital investment, to the tune of billions of dollars.” Mem. Op. at 157 (emphasis added); see also Id. at 158 (“a world class search engine is at least a $2–4B/year R&D investment”) (quoting UPX266 at 986). And handing Google’s billion-dollar know-how to competitors would dramatically reduce their incentive to pursue their own approaches to search engine innovation. Confiscating the fruits of Google’s investments? would not just impair and disincentivize Google’s ongoing development of its GSE and other assets: It risks chilling innovation industry-wide. See, e.g., Hovenkamp, supra, at 68.
Although limited duties to deal may be imposed under special circumstances “at or near the outer boundary of Section 2 liability,” Trinko, 540 U.S. at 409, an order compelling Google to share—and in some cases reconfigure—such a broad range information resources, and to do so on a long-term basis, would entangle this Court and a hypothetical “technical committee” in perpetual regulation. The diverse and sweeping sharing obligations described in sections VI, VII, and VIII of the Plaintiffs’ Proposed Judgment are not simple, straightforward, or innocuous. Indeed, the technical complexity of ongoing real-time search index exchange dwarfs that of servicing the specific operations support systems that were at issue in Trinko, and would require the court to engage in de facto regulation of, among other things, “compute” costs, APIs, API bandwidth, refresh cycles, error correction, and much more. This is a far cry from a “limited” duty to deal.
Courts are ill-equipped to manage such arrangements. Individually and collectively, the proposed sharing mandates seem tailor-made to illustrate the Supreme Court’s general caution against enforced sharing as they would require “courts to act as central planners, identifying the proper price, quantity, and other terms of dealing—a role for which they are ill-suited.” Trinko, 540 U.S. at 408.
The Court has already acknowledged these concerns in dismissing the Plaintiff States’ SA360 “refusal to deal” claim against Google, finding: “Plaintiff States seek to bypass the ‘no duty to deal’ doctrine entirely. . . . Adjudicating Plaintiff States’ claim would require the court to act as a ‘central planner’. . . grappling with a host of questions that the court is ill-equipped to handle. . . . And those thorny questions foreshadow the challenges the court would face in administering a remedy.” Mem. Op at 268-69.
A. AI Innovation and Other Technological Changes Undermine the Claimed Competitive Benefits of Proposed Data-Sharing Remedies
These difficulties are magnified in markets characterized by rapid technological change. Most of the growth in the tech sector comes “from innovation, not from increased competition under constant technology.” Hovenkamp, supra, at 68. Hence, the rapid pace of innovation, both in and adjacent to the market, “invites extra caution about intervention. While opportunities for harm might be greater, efficiencies and other positive effects are greater as well.” Id.
This Court found that “the advent of artificial intelligence (AI) has not sufficiently eroded barriers to entry—at least not yet.” Mem. Op. at 163. That uncertainty led the Court to discount the competitive significance of AI-powered search in finding liability, as “[n]ew technologies may lower, or even demolish, barriers to entry, but such innovation is meaningful only if it can change the market dynamic in the ‘foreseeable future.’” Id. (quoting Microsoft, 253 F.3d at 55). But that conclusion was based on findings of fact and trial evidence from over a year ago: AI has progressed spectacularly since then. And foreseeability cuts the other way when courts are tasked with fashioning forward-looking remedies to ameliorate harmful conduct. Courts cannot tailor remedies that are to run five or ten years based on what they cannot foresee.
Indeed, courts should evince a special degree of interventional humility given the dynamic nature of the markets at issue, including the rapid pace (and diversity) of innovation in artificial intelligence. As the Court recognized, AI is increasingly important to established GSEs, see Mem. Op. at 40-42, and central to the design of generative-AI-based GSE entrants such as Perplexity, SearchGPT, and Claude (the last of which incorporated a search API just this week, see Introducing Web Search on the Anthropic API (May 7, 2025), available at https://www.anthropic.com/news/web-search-api). Calendar year 2024 saw tremendous gains in AI performance along various dimensions, including rapid gains from open-weight models and smaller models. See generally Stanford Univ. Inst. For Human-Centered Artificial Intelligence, Artificial Intelligence Index Report 2025, at ch. 2, available at https://hai-production.s3.amazonaws.com/files/hai_ai_index_report_2025.pdf. Fundamentally, we do not know what GSEs will look like, or whether GSE’s will comprise a distinct, competition-relevant market, five or ten years hence.
That progress directly implicates questions about the magnitude and likely durability of presumed scale, scope, and quality advantages in existing data sources, as well as the magnitude and likely durability of traditional GSE competitors’ compute advantages. These, in turn, confound the larger question of entry barriers and Plaintiffs’ assertion that access to current Google search data is essential to competition. It is by no means clear what sorts of data might be required for entrants or incumbents to achieve minimum efficient scale as AI continues to improve. And in rapidly changing technology markets, it is a mistake to assume that “the successful way something has been done, and is done today, is the only way to do it, or the only way new entrants can do it and be competitive. … [I]nnovation has never required implementation of the same business model as incumbents, and especially not access to the particular proprietary inputs incumbents have created.” Geoffrey A. Manne & Dirk Auer, Antitrust Dystopia and Antitrust Nostalgia: Alarmist Theories of Harm in Digital Markets and Their Origins, 28 Geo. Mason L. Rev. 1279, 1395-96 (2021).
Note too, that the market definition that played a key role in analyzing market power for liability purposes may be confined to a more limited role at the remedies stage. As the Court observed, sources of data that are useful as GSE production inputs extend well beyond the GSE market itself already. Hence, “Microsoft has partnered with more than 100 providers to obtain structured data, and those partners include information sources like Fandango, Glassdoor, IMDb, Pinterest, Spotify, and more.” Mem. Op. at 20-21¶ 47. By the same token, even if specialized search providers may not be in the relevant market for purposes of assessing liability, they may nevertheless be quite relevant to determining appropriate remedies as their data may be used by GSE competitors. See, e.g. Perplexity: AI Search Disruptor Is Taking a Big Swing with New Shopping Feature, adepto.com (Nov. 19, 2024), available at https://addepto.com/blog/perplexity-ai-search-diruptor-is-taking-a-big-swing-with-new-shopping-feature/ (“Perplexity’s integration with Shopify allows it to tap into a vast network of merchants, enhancing its product offerings while maintaining a user-friendly interface.”).
B. The Proposed Data-Sharing Remedies Also Entail Significant Costs Arising from the Risks to Consumer Privacy and Data Security
The problems inherent in a forced data-sharing remedy are not confined to uncertainties regarding future competitive dynamics; they also implicate very real present-day risks to user privacy.
Plaintiffs’ Proposed Judgment would require Google to “use ordinary course techniques to remove any Personally Identifiable Information” from user-side data in diverse data sets, and disclose, among other things, “a description of what the dataset contains such that it can be reasonably understood by the Qualified Competitors, including but not limited to a description of what the dataset contains, any sampling methodology used to create the dataset, and any anonymization or privacy-enhancing technique that was applied” to each data set. ECF No. 1184-1 at 17-18. All of that is to be done within six months. Yet Plaintiffs provide no analysis of the implementation that would be required of Google or the direct and indirect compliance costs. Plaintiffs assure the Court that the mandated sharing provisions they seek have the advantage of “safeguarding personal privacy and security.” Id. at 16. But that assurance is hollow given the remaking of Google’s data usage that would be involved. Google’s internal use of diverse data sets comprising or derived from consumer data—some documented and some deeply embedded in trained models— does not, as a general matter, entail the same privacy and security risks as the sharing of such data with third parties—practices which are not generally part of Google’s business model. Plaintiffs do not address the complexity, cost, or even tractability of the overhaul that data sharing would impose upon Google’s data collection and use practices and, in turn, on Google’s products.
Moreover, no such reworking of Google’s internal data can eliminate the risks to privacy and data security implicated in data sharing with Qualified Competitors, as well as downstream collection, use, and sharing by other third parties. The limited assurances Plaintiffs would require of “Qualified Competitors” and the hypothetical data security standards to be developed by the Technical Committee provide only untested and conclusory assurances that downstream risks to consumers will be minimized. The last thing a remedy should do is inadvertently weaken data security or user privacy protections in the name of helping a competitor.
IV. Trinko Applies to Remedies, Not Just Liability
While modern antitrust litigation sometimes separates liability and remedy phases, the logic underpinning Trinko’s skepticism applies directly to remedies, as it is inherently focused on the practical consequences and administrability of judicial intervention. Indeed, “the difficulty of providing an appropriate antitrust remedy was central to the Trinko Court’s” holding. Barnett, supra, at 33. Furthermore, the procedural separation of liability and remedy in this case, combined with the specific nature of the liability findings, underscores the need for remedial caution.
The Department of Justice has asserted elsewhere that “Trinko addresses antitrust liability, not remedies.” Brief of the United States of America and the Federal Trade Commission as Amici Curiae in Support of Neither Party on Google’s Motion for Stay, Epic Games v. Google, Case No. 24-6274, at 6 (9th Cir., Oct. 26, 2024), available at https://www.justice.gov/atr/media/1375086/dl. Yet the Supreme Court disagrees and, as recently as 2021, relied extensively on Trinko in explaining why courts must be cautious in fashioning antitrust remedies. See Nat’l Collegiate Athletic Ass’n v. Alston, 594 U.S. 69, 102 (2021) (warning about remedies that “could wind up impairing rather than enhancing competition”, impose costs that “exceed efficiencies gained,” and “suppress procompetitive innovation”) (citing Trinko, 540 U.S. at 415).
The concerns animating the Trinko Court—especially the risk of chilling innovation incentives and the institutional limitations of courts acting as “central planners”—are not merely abstract principles relevant only to liability determinations; they are practical challenges implicated by the imposition of certain remedies. Thus, as this Court acknowledged in dismissing Plaintiff State’s SA360 claim, while adjudicating a duty-to-deal claim implicates “a host of questions that the court is ill-equipped to handle,” Mem. Op. at 268, so, too, do “those thorny questions foreshadow the challenges the court would face in administering a remedy.” Mem. Op. at 269 (citing Trinko, 540 U.S. at 415) (emphasis added).
Indeed, the disincentive for a firm to invest in potentially valuable assets or facilities arises most acutely not from a finding of liability per se, but from the subsequent remedial obligation to share those assets with competitors on terms dictated by a court. Similarly, the “central planner” problem identified in Trinko is fundamentally a remedial concern, highlighting the profound difficulties courts face when tasked with designing, implementing, and overseeing complex arrangements of forced cooperation or access.
Plaintiffs chose to pursue liability based significantly on inferences drawn from Google’s search distribution agreements and their purported effect on user choice and rival scale. Having asked the Court to premise liability on that basis, Plaintiffs cannot now escape the consequences of their chosen path. As a result, the scope of permissible remedies is necessarily constrained by the absence of but-for causation linking Google’s search distribution agreements to the foreclosure of competition. None of this changes with the bifurcation of the liability and remedies trials; rather, remedies must remain strictly proportional to, and causally linked with, the specific violations proven in the liability phase.
CONCLUSION
For the reasons set forth above, amici urge the Court to reject the far-reaching remedies of Plaintiffs’ Revised Proposed Final Judgment. Instead, remedies should be narrowly tailored to address the specific conduct found to violate Section 2, and they should be calibrated to remedy harm properly attributed to that conduct without harming competition and consumers.
[1] No counsel for a party to this case authored this brief in whole or in part. No party or counsel for a party contributed money that was intended to fund preparing or submitting the brief, and no person other than amici or amici counsel contributed money that was intended to fund preparing or submitting the brief.