The DOJ Gets AI and Antitrust Wrong in Google Case

By Ethan Yang. He is a Legal Associate at the Cato Institute and an Adjunct Fellow at AIER.

"After prevailing at the district court level in its antitrust case against Google, the Department of Justice (DOJ) is now seeking to break up the company, not just to protect rival search engines, but to protect the artificial intelligence (AI) market. Now that the matter is at the remedies stage, the DOJ argued that [t]his court’s remedy should be forward-looking and not ignore what is on the horizon.” Aside from asking the Court to speculate about possible markets that currently don’t exist, the DOJ was primarily referring to the ongoing AI race. 

However, the DOJ’s belief that Google must be torn apart to protect AI innovation is not only premature, it rests on tired misconceptions about the role of antitrust in this country’s economic history. Hampering Google for the sake of helping its competitors will undermine economic dynamism, not enable it. 

Background

In 2019, the DOJ launched a lawsuit against Google challenging certain practices related to its search engine under Section 2 of the Sherman Act. DOJ argued that Google’s contracts with internet browser companies, such as Apple, which operates Safari, to make Google Search the default search engine, denied competitors such as Microsoft Bing meaningful exposure to users. The district court agreed with the DOJ and ruled that Google’s conduct was illegal. 

However, the decision was controversial, with Google and many in the antitrust community arguing that Google’s success in the market came from its superior user experience compared to competitors. Furthermore, Google’s contracts with internet browsers further consumer welfare by increasing convenience and improving the quality of search results in a market filled with highly capable competitors like Microsoft. 

Although Google is expected to appeal the decision, the parties are currently discussing potential remedies, which the DOJ argues must involve structural separation. Although this would be an extreme idea, the agency asserts that nothing else can correct for the tech platform’s massive advantage over competitors after years of being the most popular search engine and the user loyalty and data it has gained. In fact, the government has suggested that Google be forced to share its search data, an essential corporate secret, with other AI developers to level the playing field. 

Intervening in an Undefined Market

Although the government’s depiction of Google as a dangerous AI monopolist is unsurprising, the market reality cannot be further from the truth. For one, competition in the AI market is intensifying as the technology is applied in different ways and firms continue to make improvements to their products. Google Gemini, its flagship generative AI product, is dwarfed by OpenAI’s Chat GPT in usage and is only the third most popular platform after Meta AI. Furthermore, the use of AI continues to expand into areas never imagined, even within the last year. 

In December of 2024, OpenAI released Sora, a model that creates video clips based on user prompts, and in February, Meta launched a collaboration with Ray-Ban to incorporate a virtual assistant into eyeglass frames. Finally, new players continue to emerge, ranging from startups to established firms launching their own AI models. 

Antitrust intervention, particularly in vertical restraint cases such as the matter of Google Search, is best done after extensive economic research and when a market is thoroughly understood. Otherwise, the government is just breaking apart and interfering with business conduct that it does not understand. However, that is precisely what the DOJ is doing by seeking remedies to address Google’s dominance in not just the search engine market but also the AI sector, which is still in its infancy. Indeed, when the DOJ sued Google in 2019, generative AI was not even on the radar. 

The DOJ’s Flawed Bottleneck Theory

The DOJ believes that Google is capable of using its dominance in the search engine market to control important inputs to develop quality AI models, namely, user data in the form of search queries. This is what is often dubbed a bottleneck, where an incumbent firm or first mover achieves an important market position that observers fear may allow it to control essential inputs towards downstream markets. In this case, search engine queries used to train AI models. 

Unfortunately, the bottleneck dilemma is a constantly recurring misconception in the history of American antitrust, and the lesson is never learned. Many credit the use of antitrust against incumbent firms as a means of dismantling bottlenecks and unleashing a wave of innovation, such as in the case of Microsoft or IBM near the end of the 20th century. However, although the timeline may be correct, in that these companies were sued by the government and rapid progress followed, antitrust was essentially hopping in front of the parade at the last minute. 

For example, in the case of US v. Microsoft, which occurred during the so-called “browser wars” during the birth of the internet era, the company occupied a similar position to what the DOJ argues Google maintains today. Microsoft had a strong position in the operating system market with Windows and was using it to monopolize the internet browser market by preinstalling Internet Explorer on the platform while making it harder to install other browsers. 

However, due to a mix of technical and business shortcomings, Internet Explorer began losing significant market share to Google’s Chrome Browser, which was far more nimble and useful. Although antitrust action forcing Microsoft to open up its operating system may have played some role in hastening Internet Explorer’s demise, its fate was inevitable. 

IBM suffered a similar fate decades earlier in its attempt to control the personal computer (PC) market. Although some may argue that IBM’s domination of the market was prevented by its fear of antitrust liability, the reality is that the company made a number of poor business decisions that ultimately allowed it to be outcompeted by more innovative firms. 

IBM was a power player in the PC market in the early 1980s, but soon lost control by the end of the decade. After initially creating some of the most popular PCs, it then attempted to maximize profits by making its system proprietary, perhaps thinking it could capitalize on its popularity and force consumers to exclusively use its products. However, this decision backfired as competitors stuck with a more open design that consumers ultimately preferred. 

The Wisdom of Verizon v. Trinko

The late Justice Antonin Scalia, in his unanimous opinion in Verizon v. Trinko, wrote that what some may call predatory or monopolistic conduct “is an important element of the free-market system. The opportunity to charge monopoly prices—at least for a short period—is what attracts “business acumen” in the first place; it induces risk taking that produces innovation and economic growth.” That is, if a company like Apple achieves a strategic position in the smartphone market, it should be able to enjoy the fruits of that achievement, not be forced to help its competitors in the name of equal competition. 

In fact, exploiting a dominant market position by charging higher prices or making more proprietary products creates greater incentives for rivals to challenge the incumbent. 

If a company like Google is able to obtain large amounts of user search data, it should be able to use it to create a powerful AI model and not be forced to share that advantage with its fiercest competitors. Indeed, it is precisely these asymmetries that the DOJ argues will destroy competition that actually encourage more vigorous innovation. When companies cannot rely on the government to force their rivals to hand over key inputs, building a better or different mousetrap is the answer. Instead of being able to copy Google’s AI model because they can just force the company to share search data, competitors must find other ways to succeed. 

The emergence of the Chinese AI, DeepSeek, is perhaps the most obvious example of Justice Scalia’s logic that monopolies refusing to aid their competition is often healthy for competition. Because the company was shut off from the talent, knowledge, and semiconductors needed to compete with American AI, the Chinese firm had to make do with less. DeepSeek, through a mix of reverse engineering American AI, experimenting with less powerful semiconductors, and shrewd talent management, produced a powerful, less costly model that took the world by surprise. So even in the face of even more daunting, government imposed barriers, DeepSeek demonstrated that mere size and resources do not determine success. 

Forcing Google to open up its system to competitors will likely decrease, not increase, competition and degrade the welfare of consumers. On the one hand, Google is deprived of the resources it worked hard to obtain while also being less incentivized to improve its products, since its competitors will be able to free-ride off the company’s efforts. Google’s rivals will feel less pressure to improve and innovate and consumers will ultimately lose in the end. 

Justice Scalia discussed the irony of such interventions in Trinko when he wrote, “[m]istaken inferences and the resulting false condemnations ‘are especially costly, because they chill the very conduct the antitrust laws are designed to protect.”

Rather than unleashing competition and innovation in the AI sector, the DOJ’s fixation on restraining Google so that its rivals may flourish will have the exact opposite effect. Overly speculative and precautionary antitrust that is more concerned with the health of other firms rather than the ultimate benefit to consumers will create an equal playing field for mediocrity, not rigorous competition."