The FTC is setting its sights on generative AI

Watershed technological transitions can usher in opportunities for new entrants to challenge market leaders. These rare paradigm shifts redefine how companies compete for customers and resources. The emergence of generative artificial intelligence (AI) is a quintessential example of how innovation can either disrupt or entrench dominant incumbents depending on how markets and regulators respond.

Generative AI uses massive models trained on rich and diverse datasets to create new content. This revolutionary tool is reshaping how businesses interact with their customers, competitors, and partners, creating immense opportunity and great risk.

The Federal Trade Commission (FTC) is making a case for aggressive antitrust enforcement. A recent Technology Blog submission from the Bureau of Competition and Office of Technology staff outlines several practices that could trigger government intervention. To understand antitrust risk when competing in markets affecting generative AI, businesses should familiarize themselves with the parameters and limits of several common antitrust theories of harm.

Exclusive dealing

Companies often seek to shore up suppliers or customers through exclusive deals. Exclusive dealing is not necessarily problematic and often stimulates competition. But agreements that enable one firm to control a critical input, distribution channel, or customer segment can raise concerns. In McWane v. FTC, for example, a federal court condemned an exclusive-dealing arrangement that foreclosed rivals from “distribution sufficient to achieve efficient scale, thereby raising costs and slowing or preventing effective entry.”

In the context of generative AI, the FTC foresees antitrust exposure where “incumbents that offer both compute services and generative AI products” wield such arrangements to discriminate against new entrants. The FTC appears poised to scrutinize exclusive deals involving compute resources, such as graphical processing units, that are key to competing for generative AI markets.

Although each agreement requires an individualized assessment, several general principles are worth knowing. First, exclusive-dealing arrangements should not be implemented as part of a scheme to exclude or deny rivals the ability to compete. Exclusive-dealing contracts that block competitors from scaling are inherently risky.

Second, companies should contemporaneously document any pro-competitive benefits, such as lower costs, higher quality, and better access to products, and should be prepared to explain specifically how the exclusivity results in improved products or services. Finally, be aware that exclusive arrangements with companies whose market share exceeds 30% are riskier.

Tying

Tying generative AI to adjacent products or services is also on the FTC’s radar. Under antitrust laws, tying involves conditioning the sale of a dominant product on the purchase of a second product. The FTC cautions that firms “may be able to link together new generative AI applications with existing core products to reduce the value of competitors’ standalone generative AI offerings, potentially distorting competition.”

As the Supreme Court explained in Jefferson Parish, an illegal tying arrangement occurs when a seller forces a buyer “into the purchase of a tied product that the buyer either did not want at all or might have preferred to purchase elsewhere on different terms.” Tying arrangements are viewed less favorably when “the seller’s share of the market is high or when the seller offers a unique product that competitors are not able to offer.” It is important to recognize that tying is generally not an antitrust issue unless the two products are “distinguishable in the eyes of buyers.”

Certain best practices can reduce the risk of antitrust liability. First, track instances where a customer expresses an affirmative desire to purchase the products together. Second, demonstrate how the arrangement benefits competition, including by reducing costs, creating efficiencies, increasing the combined value of the tied products, and enabling technological improvements.

Bundling

As with tying, bundling leverages a highly desired product to increase sales of additional products. In Cascade Health v. Peacehealth, the bellwether U.S. Court of Appeals for the Ninth Circuit Court, bundling is described as “the practice of offering, for a single price, two or more goods or services that could be sold separately” for a lower combined price. Bundling often results in significant consumer savings, but in theory it could also hinder “a potential competitor who does not manufacture an equally diverse group of products.” Courts are skeptical when sellers with 30% or greater market share offer bundles that include products priced below cost.

The FTC points out that companies could conceivably bundle productivity software, web browsers, or cloud services with generative AI to establish a broad user base. In their view, such conduct could prevent less-established generative AI firms from gaining a foothold in the market. To assess antitrust risk, companies should consider whether the breadth of products in the bundle prevents rivals that are offering a narrow suite of products from competing effectively.

Acquisitions

Enforcers are also on the lookout for acquisitions of key inputs to the large models that feed generative AI offerings. This includes vast datasets, specialized engineering talent, and state-of-the-art computational power. The agency is sure to pay close attention to M&A in these markets, including nonhorizontal mergers and acquisitions.

The FTC’s policy of “vigorous law enforcement” in this space is a clear signal that antitrust regulators will be closely scrutinizing conduct affecting generative AI. Schemes crafted to eliminate, harm, or exclude competition, on the other hand, will likely invite a vigorous regulatory response. Companies attempting to corner markets through anticompetitive conduct risk costly and time-consuming investigations and litigation.

To mitigate antitrust risk, businesses should focus on how exclusive deals, tying, bundling, and strategic acquisitions are necessary to compete effectively through higher-quality and more cost-effective AI tools.