Fiona Scott Morton and David Dinielli show how landmark antitrust cases historically have cleared the path for innovation in the next “frontier technology.” But with closing arguments in the search monopoly case just days away, Google threatens to evade this round of rigorous new competition. It reportedly is in talks to place its own artificial intelligence on Apple devices as it did in the case of search. Such a maneuver would entrench Google’s search monopoly and place Google in the driver’s seat to steer the development of consumer-facing AI. The authors offer up a menu of steps the government might take now to thwart Google’s new anticompetitive strategy and preserve competition in AI before it’s too late.
The cases in which the United States Department of Justice and a coalition of states challenge Google’s alleged illegal monopoly in search will conclude in early May with closing arguments on the question whether Google broke the law. The government case looks strong and the plaintiffs may well win on the merits. But a legal victory will not matter if Google has outmaneuvered enforcers to extend its monopoly into the market for the next important industry: artificial intelligence-enabled consumer tools. Public reports indicate that Google is negotiating a contract with Apple to put Google’s generative AI chatbot, Gemini, on the iPhone. This could be a repeat of the anticompetitive playbook Google used in search, where it paid Apple to set Google Search as the default search engine on its Safari browser. If Google is similarly successful, this new move could permit the company to entrench and extend its search monopoly, frustrating the last five years’ worth of antitrust investigation and enforcement designed to break Google’s grip on search. It also could starve consumers of the benefits of rigorous competition in AI, condemning them instead to an AI future that Google—the alleged monopolist—decides is best for Google.
Consider the lessons from the long tradition of federal antitrust enforcement in frontier information and telecommunication technology, recently recounted in Giovanna Massarotto’s “Driving Innovation with Antitrust,” here in ProMarket. The AT&T settlement in 1956 prohibited AT&T from entering the computer industry despite its innovations like Unix. The resulting level playing field allowed the computer hardware industry to flourish and gave us IBM. The government’s antitrust case of 1969 against IBM caused IBM to separate mainframe hardware and software. This allowed the software industry to flourish and gave us Microsoft. The DOJ’s antitrust case against Microsoft was decided in 2001 and prevented Microsoft from controlling the development of the internet. Specifically, the decree entered in that case prohibited Microsoft from disadvantaging its oblique competitors: those that competed with Microsoft not directly, but by developing new tools and services such as search. The chief beneficiary of the Microsoft decree, it turns out, was Google.
Today, we have active litigation against Google in search; this litigation, assuming it is successful, should put a halt to Google’s maintenance of that monopoly, as well as allow competitive development of the next new frontier technology. A principal objective and effect of these historic monopolization cases was to open up the path of future innovation. Because innovation is inherently an uncertain process, we cannot be sure what will come next, but some form of consumer-facing artificial intelligence seems likely. What we do know, however, is that society is made worse off when the incumbent monopolist seizes control of the path of innovation and excludes others who might offer a more exciting direction, product, or idea. We should expect no less from the current search litigation than its historic precedents delivered: the fencing in of the current monopolist to make space for innovation in the next market.
Some commentators, perhaps unaware of the enforcement history described above, have set their sights lower, and have argued that we should not be concerned about Google’s market power in search because new innovations will come along and disrupt it. Ironically, this perspective most often identifies AI as that threatening new innovation.
We do not doubt AI’s disruptive potential. Indeed a recent report from the United Kingdom’s Competition & Markets Authority (CMA) identifies a handful of large language foundation models that compete head to head with Open AI’s Chat GPT 4, including Google’s Gemini Ultra and Anthropic’s Claude 3. Despite the inter-firm rivalries, however, the same CMA report identifies several significant threats to competition in this new field and related markets, including the possibility that (1) incumbents might use their positions in consumer- or business-facing products to foreclose competition in AI; and/or (2) partnerships with key AI players could extend or reinforce positions of market power throughout the value chain. These are precisely the concerns posed by the reported talks between Google and Apple, which could deliver to Google the data and usage of two billion active Apple users. History shows, and the CMA recognizes, that monopolists do not ride off into the sunset voluntarily. Indeed, reports about a new agreement between Apple and Google indicate Google may intend to do the opposite.
Although details of the negotiations remain under wraps, it’s not difficult to imagine how an agreement between the two companies might help to maintain Google’s search monopoly in a way that frustrates competition and innovation in AI, particularly if Apple has agreed to give Google AI products exclusive default positioning on Apple devices, as it did with Google search. Generative AI can supplement a traditional search engine. Bing’s Copilot feature is the prime example. From the consumer standpoint, the AI simply complements the traditional service. Absent an agreement, we could expect consumer demand to force Apple to develop its own AI for use in searches or offer one of the other third-party options. But it appears, based on the public reporting, that Apple may shortcut that path and simply install the Google products instead. As the New York Times reported in late March:
In a note on Tuesday, a Bernstein Research analyst, Toni Sacconaghi, called an Apple-Google deal a “win-win,” giving Apple generative A.I. for iPhones and validating Google’s work on Gemini. He also said Apple didn’t have to own an A.I. model on iPhones to profit from it and could instead take a commission from Google, which currently charges $19.99 per month for its Gemini Advanced app.
This would be good news indeed for Google, which wants to keep all users searching on Google, whether through its traditional search engine or a new AI complement. It would allow Google to further maintain its monopoly power in the markets alleged in the search cases. If Apple and Google already have agreed to place Google’s AI product on the Apple iPhone as an exclusive default, then it will become the default choice for almost 100% of handset users in the U.S. and Europe—given that Google is highly likely to make its own AI product the default on Google Android handsets also. A remarkable aspect of this development is that the parties appear to be negotiating—or may have negotiated—a contract to keep rivals out of AI-enabled mobile search during the pendency of an antitrust case where the government has accused Google of blocking entry by rivals in mobile search.
Can the U.S. government do anything to make its enforcement efforts of the last five years generate some benefit for consumers? The court has scheduled closing arguments (Dkt. 890) in the current Google Search case for May 2-3, just one week away. Prior to those closing arguments, or immediately thereafter, the trial teams could use motion practice to raise this issue with the court. In particular, the court might be interested to learn that its effort to hear and rule on the case will largely be wasted, even if it rules in favor of the government, because the defendant has used the same tactic again in order to leverage into a new monopoly position.
Ideally, the closing arguments themselves would provide a venue to raise, explain, and provide evidence of the evasive strategy. The problem is that such a discussion requires evidence. That evidence is lacking because this recent strategy was not, and could not have been, described in the original case filed in 2020. Enforcers have not taken relevant discovery and, of course, presented no evidence about the strategy at trial.
Google must not be allowed to evade or delay scrutiny of its new strategy simply by birthing it post-trial, however. Rather, enforcers should take immediate action in connection with the current search case with an aim to accomplish two related goals. First, enforcers should act aggressively to place on hold any agreement between Apple and Google regarding AI until it can perform a factually informed analysis of whether the effect of the agreement would be to maintain Google’s search monopoly or otherwise harm competition and consumers. Second, enforcers should take immediate action demonstrating to the court and to the public a serious effort to preserve the court’s ability, if it finds liability, to craft a full remedy—one that addresses Google’s current efforts to maintain its monopoly and not just its past misdeeds. Steps enforcers could take immediately include the following:
⇒ The government could ask Google to agree to modify the Stipulated Protective Order (Dkt. 86 at 4-6) to require preservation of all evidence regarding the ongoing negotiations. If Google refuses, the government should file a motion that explains the importance of that evidence and asks the court to require that Google preserve it.[1]
⇒ The government could, pursuant to Rule 65, ask for a temporary restraining order and then a preliminary injunction to preserve the status quo from now until the entry of judgment, preventing the consummation of any new agreement of the sort press reports describe. Because a change to the status quo could lead to irreparable harm to competition—harm that a remedy decided in a later phase could not fix because the AI market had evolved in response to the Google-Apple agreement—a preliminary injunction could protect competition.[2]
⇒ The government could petition the court under Rule 27 for an order requiring Google to produce certain evidence and to permit depositions relevant to the negotiations. The verified petition, which would be heard on 21 days’ notice, would alert the court to the competition concerns posed by the negotiations and, if granted, generate admissible evidence the court could consider in ruling on the merits of the search case.[3]
There are longer-run solutions that could also be deployed. The government could bring Apple into the process by amending the recent Apple complaint alleging monopolization of smartphone markets to include the agreement with Google, naming both Google and Apple as defendants in a Section 1 count. However, this case will not be tried for three or four years if the last digital cases are any guide, and then it may be appealed, by which time competition in AI may be permanently lessened.
Assuming the court finds liability in the Google Search case, the government may be permitted to take discovery about any Google-Apple agreement regarding AI, and could incorporate the resulting understanding into the remedy it proposes. One such remedy could be a permanent restriction on any such agreement. A broader option is a permanent injunction or consent decree that prohibits or limits Google’s activities in AI. In substance, the proposed remedy might resemble the 1956 AT&T consent decree prohibiting AT&T from entering computing. A weaker option is a mandatory separation between Google’s AI businesses and search. If a court chooses to separate Google’s activities, significant internal reorganization of the company will be required. Divisions must be run separately, accounts calculated separately, employees may not be transferred between divisions, services must be purchased using arm’s length prices and terms, etc. In particular, such an AI division should have no contract with Apple for a fixed period of years to open up the AI space to other rivals. The downside to each of these ex post solutions, of course, is that the remedy must put Humpty Dumpty back together again rather than stopping him from breaking in the first place.
Finally, enforcers could use public speeches to explain the need to keep AI competitive by prohibiting the repetition of the tactics that were so successful in search. Sending a clear message that enforcers are already linking the corporation’s behavior across technologies may cause Google to weigh the cost of harsher remedies, a second antitrust case, or a court that views the corporation as a repeat offender. But regardless of what tactic enforcers use, they must not sit idly by at this critical moment. Consumers will lose if Google’s monopoly, unlike those of other tech giants that preceded it, allows it to control the course of future innovation.
[1] Enforcers could bring the motion pursuant to Rule 16(b)(4) of the Federal Rules of Civil Procedure, which provides for the modification of a pre-trial order’s scheduling provisions—provisions that Rule 16(b)(3)(B)(iii) expressly confirms may address the parties’ obligations to preserve ESI, for example.
[2] Plaintiffs typically request a temporary restraining order and/or preliminary injunction at the time they file a complaint, but such a request is proper at any time. Indeed, Google likely will respond that the issue is not even ripe. But, generally, sooner is better when it comes to requests for injunctions, especially given the traditional preference in equity for prohibitory over mandatory injunctions. Indeed, prohibitory injunctions often issue to preserve the status quo when, as is the case here, a change to the status quo could frustrate the possibility of awarding complete relief. Google may argue that it stands to lose vast sums if its deal is blocked even temporarily, but the argument is no reason not to enjoin consummation of the agreement. First, it is expected that monopolists will lose revenue and income when forced to stop monopolizing. Second, no matter how much Google claims it might lose, Rule 65(c) provides that the United States needn’t post security as a precondition to the issuance of an injunction.
[3] Rule 27 authorizes a person who expects to be a party to anticipated litigation to conduct a “deposition to perpetuate testimony.” Assuming the petitioner meets the rule’s prerequisites, and despite the rule’s language and title, courts have read Rule 27 broadly to authorize not only a deposition, but also orders for the pre-lawsuit production of documents, ESI, or physical evidence. See generally Depositions to Perpetuate Testimony, 8A Fed. Prac. & Proc. Civ. CIV Rule 27 (3d ed.) Necessary averments would include (1) that the government anticipates litigation regarding the Google/Apple agreement but cannot presently file suit (for example because of internal processes that require certain investigative and/or approval procedures); and (2) reasons for the need to “perpetuate” the evidence (such as the need to obtain contemporaneous evidence of the current status of negotiations, and/or Google’s demonstrated negligent failure in the search case itself to comply with preservation obligations). A Rule 27 petition is to be filed as a miscellaneous matter that likely would be deemed related to the search case pursuant to Local Civil Rule 40.5(a)(3). Once the petition is filed, it would be an easy lift to ask the court during closing argument to take judicial notice of the Rule 27 petition. If the court were to grant the petition, any consequent depositions taken of key Google witnesses would be admissible against Google as admissions.
Authors’ Disclosures:
Fiona Scott Morton is the Theodore Nierenberg Professor of Economics, Yale School of Management; Professor (Adjunct) of Law, Yale Law School; Research Fellow, Tobin Center for Economic Policy, Yale University. She regularly consults for governments and private parties on antitrust matters. Corporate clients in the last three years include Microsoft and Amazon as well as a few healthcare corporations and consumer goods providers.
David Dinielli is a Visiting Clinical Lecturer in Law & Senior Research Scholar, Yale Law School; Senior Policy Fellow, Tobin Center for Economic Policy, Yale University. In 2011-12 he was Special Counsel at the U.S. Department of Justice, Antitrust Division. He has no conflicts of interest to disclose.
The authors state that they have received no financial support for the preparation of this essay and that no person other than the authors had the right to review this essay prior to its circulation.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
