Competition Corner: Maximum DMAge
Issue #44
G’day,
This month saw the release of President Trump’s AI Action Plan and (as Alden Abbott writes) it’s broadly great news for fans of competition and innovation. The plan outlines a vision of “permissionless innovation,” eschewing proscriptive and burdensome regulations in favor of letting businesses and innovators experiment, develop and grow. It promises to slash red tape and help AI businesses and workers through tax concessions and public investments in datasets, labs and other resources. It also threatens states that unduly hamper American AI innovation through burdensome regulations with funding cuts. This will hopefully help ensure that we don’t end up with a grim patchwork of differing laws across jurisdictions that would hamstring cross-border data flows and AI-facilitated interstate commerce, as I’ve written previously.
American AI innovation is being driven by both burgeoning startups as well as large digital platforms that both collaborate and compete inside a dynamic competitive ecosystem, as Alden Abbott and I noted in a recent Mercatus Center working paper. But it isn’t just threatened by rogue state governments. It’s also threatened by foreign ones seeking to impose proscriptive ex-ante regulatory frameworks that unduly single out and target American firms in the name of policing uncertain and spurious harms.
South Korea, Thailand, Columbia, Japan and Brazil are all considering their own versions of the European Union’s recently implemented Digital Markets Act (DMA). The DMA singles out a narrow list of firms (almost all of whom happen to be American tech giants or digital platforms) for a range of proscriptive mandates. These restrict their business practices, mandate sharing of resource access and user data with third parties to facilitate ‘interoperability,’ and threaten them with fines for non-compliance.
By singling out American digital platforms, the DMA undermines both the pro-competitive goals of the administration’s AI Action Plan, as well as the competitive prospects of European AI startups in a number of ways:
By restricting platforms like Amazon, Meta and Google from self-preferencing and integrating or bundling products with their other products, consumers are harmed through degraded user experiences. This deters them from using the platforms or fragments consumers across platforms, thereby reducing access to aggregated user data necessary for training and developing better AI products. For instance, preventing Amazon from prominently displaying its own versions of a product (“self-preferencing”) deters it from developing cost-effective competing products to popular ones on its marketplace. This, in-turn, harms the platform’s ability to develop AI tools using the sales data of its own products across a dynamic portfolio of categories to predict what consumers are likely to look for and to tailor offerings accordingly.
Uncertainty around whether specific business practices would attract a fine, prohibition (or both) under the DMA deters firms from experimenting with new AI tools or introducing them to Europe at the same time they’re made available to the rest of the world. This has already resulted in slower software updates, meaning delays in product improvements.
The threat of sweeping ‘interoperability mandates’ has similarly caused delays in allowing US firms’ AI tools to reach and benefit European consumers. For instance, the iOS feature Apple Intelligence seeks to mitigate privacy concerns around personal or proprietary user data by processing this data on private cloud servers or the individual’s own iOS devices. Uncertainty around whether this data would attract an interoperability mandate that would have curtailed data security meant that the feature and other improvements were rolled out to Europeans later than they were made available in the United States. Mandated sharing of data from search engines and other tools (including search indexing data) also discourages the development of tools that are most effective at attracting useful and comprehensive user datasets. These are essential for training the best AI foundation models and services. The tradeoff between mandating that a firm share its proprietary resources with rivals and incentives to create and make use of such valuable resources in the first place was recognized by the late Justice Scalia as far back as 2004 in the landmark Trinko case. In the long run, the disincentives to develop and invest in products that effectively attract users and aggregate their data could outweigh any positive impact from a platform’s competitors gaining access to its existing data.
Mandated “sideloading,” which restricts Google and Apple’s ability to impose data security standards for third-party apps in their app stores, can weaken features including privacy, fraud prevention and standardized ratings. This makes third-party apps less trustworthy for users, thereby restricting rather than helping many of them to attract training data while degrading and limiting their AI tools’ ability to reach users. This disproportionately impacts app developers that lack an established reputation for security and privacy. Ironically, users are more likely to choose apps offered by established incumbents due to their reputation for security, thereby helping to entrench the market share of incumbents. As things stand, competing mobile platforms already offer varying degrees of sideloading and security features. This is a tradeoff that consumers ought to face. Smooth interoperability and “open access” to efficient distribution channels for apps can boost competition. However, mandating a one-size-fits-all approach for app stores that degrades security in favor of interoperability risks compromising competition and innovation instead of fostering it.
Restrictions on platforms’ ability to offer tailored advertising to consumers harms both the consumers (who are denied access to products tailored to their preferences) as well as startups without established reputations that rely on such advertising to reach new consumers. Importantly, it also degrades user experience on the platforms and undermines the ability to develop and monetize AI tools. This discourages firms (including the platforms themselves) from investing in improving these tools or producing new ones. A recent European Commission ruling against Meta recognized that this prohibition would degrade the platform’s product offering, but held that this is irrelevant to the analysis that the DMA requires to determine a violation.
So what can the Trump administration do to address these problems and the threats that similar proposals pose, in order to keep America’s tech and AI sector vibrant, innovative and competitive?
As trade deal negotiations with the rest of the world continue, the administration can lobby other states to repeal Anti-Competitive Market Distortions (ACMDs) affecting US firms. These include the ACMDs embedded in the DMA, which favor neither the American nor the European tech and AI sectors. Instead, they threaten to cede the advantage in the AI race to China, with Bytedance being the only Chinese tech giant captured by the DMA. For more on ACMDs and how they can do even more damage to American companies than other countries’ tariff barriers, check out Alden’s writings here.
The Trump administration should also eschew bipartisan proposals to enact DMA-style ex-ante rules that are akin to the DMA, such as the proposed American Innovation and Choice Online Act and the Open Markets Act. As I have noted at length, existing US antitrust laws provide a flexible and pragmatic framework for capturing anticompetitive conduct that is adaptive to technological change. For instance, the Microsoft framework has served well in providing guiding principles for adjudicating anticompetitive conduct in digital markets. By contrast, the European Union’s embrace of restrictive ex ante regulations that restrict firms’ abilities to innovate and grow, as well as their creation of an unfavorable regulatory climate for tech mergers and acquisitions, have stymied startups’ ability to attract investment. This has also limited opportunities for founders to ‘exit’ the market, thus discouraging them from going into business or product development independently. It's no wonder then that United States firms have obtained more than six times the amount of AI investment as their European and British counterparts did between 2013 and 2023.
Finally, the competition enforcement agencies, such as the FTC and DOJ antitrust division, should reconsider their requests for remedies in cases such as Google Search, which could have detrimental effects on competition that mirror or even exceed those in in the DMA. A structural breakup of Google (which would destroy key features that consumers value by destroying the integrative synergies of the Chrome browser, search engine and other user-facing products, all of which facilitate rollout and training of AI tools), for instance. For more on the Google Search decision that is currently being appealed, check out my analysis here.
But what do you think? Let us know!
Competition Enforcers Are Coming for Cloud Computing
The Organization for Economic Cooperation and Development (OECD) recently followed the UK Competition and Markets Authority (CMA) in issuing a policy paper alleging risks of increased market concentration and anticompetitive harms in the rapidly growing cloud computing services sector. Like the CMA, they suggest a range of government interventions and new ex ante rules for addressing the concerns they raise. However, these are more likely to do harm than good. Existing American antitrust law already provides ample grounds for prosecuting credible allegations of anticompetitive business practices. This is a preferable alternative to proscriptive restrictions on business practices with both potential pro- and anti-competitive implications, as Mercatus intern Jack Trotter and I argue in Truth on the Market.
It's time to revise those Merger Guidelines
The retention of the Biden administration’s 2023 FTC-DOJ joint merger guidelines has created a conundrum for the Trump FTC and DOJ. Although the agencies emphasize their dedication to reducing merger-related government burdens (such as through the reinstitution of early termination), the 2023 guidelines continue to impose costly uncertainty that may discourage efficient mergers. Alden Abbott suggests targeted fixes to the guidelines (or at the very least clarifying speeches) in a recent Forbes article.
Intellectual Property & Competition
In my amicus brief filed before the United States Court of Appeals for the Federal Circuit, in the matter of SAP America Inc., I argue that the US Patent and Trademark Office (USPTO) director has the power to discretionarily deny an inter-partes patent review application on the grounds that the same patent is being contested in the federal court even if interim guidance issued by the previous director stated that they would not deny it at the time of the application’s filing. Though dealing with niche matters of administrative and patent law, the matter has ramifications for the costs and certainty around holding and defending patent rights, which in turn impact the competitiveness and dynamism of the United States’ innovation ecosystem.
See you next time!
Satya Marar
Visiting Postgraduate Fellow
Mercatus Center at George Mason University
 | A guest post by
|