On November 20, the Department of Justice and Plaintiff States filed a Proposed Final Judgment (PFJ) in the Google Search case, articulating the remedies they will seek. Those proposals have made waves for their arguable severity, including by proposing a divestiture of Google’s web browser, Chrome. But the PFJ has raised eyebrows for another reason; although the case concerned Google’s monopoly over search, enforcers have targeted a different, emerging market—artificial intelligence.
The Case against Google
In August, Judge Mehta issued his landmark Google Search liability ruling. In his opinion, the Court found that Google had monopoly power in general search and search-adjacent text advertising. The Court also held that Google engaged in anticompetitive tactics to obtain and protect that dominance; these included agreements with distributors (such as Apple, Mozilla, and cell phone carriers) to make Google Search a default, which illegally entrenched Google’s monopoly power.
Now the case has entered the remedies phase—and enforcers have come out swinging. In addition to enjoining the agreements directly at issue, the government plaintiffs seek more fundamental relief, to cut off Google’s monopoly power directly. On this point, the PFJ includes requirements to force Google to share key data inputs (including its famous search index), to ban payments to distributors to incentive traffic to Google, and to divest Chrome (and possibly Android).
These more fundamental reforms seek to stem the source of Google’s monopoly power—it’s flywheel. The more queries Google Search receives, the more data it has to train its product. This feedback effect results in more customers choosing Google for their queries, feeding even more data for its training. Thus, Google’s monopoly power reinforces itself, and rivals can’t get a toehold. The PFJ seeks to disrupt this cycle, by giving rivals improved access to critical data. By depriving Google of unique access to inputs, the government plaintiffs hope to allow other rivals to become viable competitors.
Artificial Intelligence: The Next Frontier
In his liability opinion, Judge Mehta considered whether emerging AI products may challenge Google’s dominance. Ultimately, the Court found that “[d]espite these recent advances, AI has not supplanted the traditional ingredients that define general search. And it is not likely to do so anytime soon.” Memorandum Opinion ¶ 114, ECF 1033, United States et al. v. Google LLC, No. 1:20-cv-03010 (D.D.C. Aug. 5, 2024). Yet in the PFJ, government enforcers have their eye on this emerging industry, worried that Google’s existing monopoly position might stunt competition in this emerging marking. Such a concern is not without precedent—landmark monopolization cases have long sought to unleash competition for products built on top of monopolized markets.
The legacies of both AT&T and Microsoft lie not only with the market that was originally dominated—telephone service, operating systems—but also with the distinct markets that were built on top of them, which the government sought to keep free from monopoly influence. For example, AT&T long dissuaded independent devices from connecting to its telephone network. This meant that for generations, only products from Western Electric (AT&T’s device manufacturer) could interconnect. Even today, the image that most people over 40 have of AT&T is a Western Electric telephone (and not its network). But once AT&T’s network was broken apart, and then separated from Western Electric, entire new devices came to market. Some of those—like the modem—came to define the next generation of innovative technology. Similar with Microsoft. In the late 90s and early 2000s, “[s]mall firms like Google, Facebook, Amazon, and others were all dependent on the web browser, over which Microsoft [then] had a monopoly”; “it is highly doubtful that Google would have achieved dominance [in search] in a world where Microsoft could dictate what search engine was being used on every computer in the world. We would all be using BING.” T. Wu, The Curse of Bigness 100 (2019).
Thus in both cases, DOJ enforcement led to the curtailing of a firm’s monopoly power in a new, emerging industry. Government enforcers are clearly worried about Google’s ability to use its existing monopoly power, which originates from its access to data inputs, to influence markets built on top of them, like generative artificial intelligence. As a result, the proposed final judgment includes several AI-related provisions, such as requiring that Google inter alia (1) disclose and divest its holdings in any AI product or another technology that is a “reasonably anticipated competitive threat[]” to general search; (2) not acquire an interest in a query-based AI product; and (3) not restrict publishers’ ability to make data available to a rival AI product developer.
In response, Google will likely argue that the PFJ’s AI-related remedies stretch too far because the liability opinion is limited to the Search and text-advertising markets. Although traditionally courts have flexibility to fashion equitable relief that extends beyond the violation, the D.C. Circuit in its landmark Microsoft decision seemed to curtail that principle, holding that relief “should be tailored to fit the wrong creating the occasion for the remedy.” United States v. Microsoft Corp., 253 F.3d 34, 107 (D.C. Cir. 2001) (en banc) (per curiam). As a result, the ability of the Court to award relief related to the AI market at all will be hotly contested. For his part, Judge Mehta has already signaled that AI “is only going to play a much larger role . . . in the remedy phase than it did the liability phase . . . Is that because of the remedies being requested? Perhaps. But is it also potentially because the market that we have all been discussing has shifted?”
Nonetheless, if imposed, these prohibitions would significantly change Google’s business strategy. Most notably, the proposal would force Google to unwind its investment in Anthropic, which has created a chatbot (Claude) to rival OpenAI’s ChatGPT. Google’s investment in Anthropic, along with other Big Tech companies’ investments in similar products, has already come under antitrust scrutiny by the FTC.
But even outside of the terms that directly affect generative AI, the PFJ’s other proposals could also dramatically shape competition going forward. Generative AI, much like search, relies heavily on data inputs to improve its product. As a result, the government’s efforts to prevent Google from having default, if not exclusive, access to queries (one source of that data) will significantly impact the emerging AI market. For example, because Chrome has 61% share of browser traffic (where the bulk of queries originate), any divestiture will mean Google will no longer have in-house, exclusive access to data from those queries. An independent Chrome could even be an important distributor for a Google rival (such as a search engine like Bing or an AI chatbot) who might purchase a default position. Further, the PFJ’s behavioral remedies related to Android and the Play Store (including their possible breakoff) means that Google will have substantially less leverage to preference its own query-based products, and/or deprive its rivals of other critical data inputs. And the PFJ’s proposed requirements to force Google to share existing data assets (including its famous search index) will benefit not only rival search engines, but also likely new generative AI products as well.
Of course, the PFJ is just a proposal. Judge Mehta’s ultimate remedies will rely on the record developed by the parties. But regardless of how the Court rules, government enforcers are clearly attentive to the emerging artificial intelligence market—in this case, and likely in future ones too.
