Locke Lord QuickStudy: Antitrust and Artificial Intelligence: ‎The FTC Identifies Competition Concerns in Generative AI ‎Markets

On June 29, 2023, the Federal Trade Commission’s Bureau of Competition and Office of Technology published a blog identifying key segments in the artificial intelligence (AI) space where competition concerns may arise.^[1] For companies buying, building, or selling products and services that incorporate generative AI, this blog provides a roadmap of markets susceptible to antitrust violations and possible unfair methods of competition.

What is Generative AI?

“Generative AI” means any system of AI that generates new content—such as text, images, audio, video, or computer code—through computer algorithms trained to learn patterns and relationships in large sets of input data. Currently there are hundreds of different generative AI products available to the public. According to the latest available data, one of the most popular generative AI tools, ChatGPT by OpenAI, has over 100 million users and its website currently generates 1.8 billion visitors per month.

What You Need to Know

• The FTC intends to protect generative AI’s potential benefits to society by facilitating competition at all levels of the AI market chain.
• The FTC specifically is concerned with “essential inputs” and “key building blocks” on which generative AI depends. These include (1) data, (2) labor talent, and (3) computational resources.
• The FTC already envisions several “possible unfair methods of competition” in the generative AI market. The FTC’s blog provides detailed examples of various activities the FTC is monitoring and may seek to deter through enforcement actions.

The FTC’s Markets to Watch

1. Data: Federal and state antitrust laws do not prohibit the collection or maintenance of large sets of data. However, a company might use its control over data (especially in highly regulated industries, such as finance or health care) to restrain fair competition.
   
   
2. Labor Talent: The FTC notes it is “critical talented individuals with innovative ideas be permitted to move freely, and, crucially, not be hindered by non-competes.” To help address this specific concern, the FTC references its January 2023 notice of proposed rulemaking that would significantly limit non-compete clauses in employee contracts (see Locke Lord QuickStudy: FTC Proposes Broad Non-Compete Ban: ‎What Employers Can Expect Next).
   
   
3. Computational Resources: Access to computing power through specialized chips (e.g., Nvidia’s H100) or cloud computing services are necessary to process large amounts of data and train models for use in generative AI applications. Because of the high barriers to entry in the markets for computational resources, these markets could become highly concentrated and their players prone to antitrust violations.

Possible Unfair Methods of Competition

1. Method: Bundling and tying.
   
   

   Examples: A company may offer a bundle of AI applications and other products together as a single package (i.e., bundling) or may condition the sale of an AI application on the purchase of separate products (i.e., tying).

2. Method: Exclusive dealing or discriminatory behavior.
   
   

   Example: A company with a large cloud computing platform may steer users toward its own AI products instead of competitors’ products or give itself preferential cloud computing services at the expense of the company’s customers.

3. Method: Mergers and acquisitions activity to consolidate market power.
   
   

   Example: A company with vast computational resources for data processing and training (i.e., thousands of highly sought after GPUs) may try to acquire companies with AI applications to prevent access by rivals.

4. Method: Leveraging the “network effects” (i.e., additional value as more people use a product or service) and “platform effects” (dependency on a platform) associated with a company’s computational assets to consolidate market power.
   
   Example: A company may offer low fees for customers to develop and run AI products using the company’s cloud computing platform and then charge the customers high fees to export the customers’ data housed in the company’s cloud.
   

Who Else is Watching?

The FTC is not the only federal agency dedicating time and resources to understanding AI tools and analyzing potential competition concerns. The AAG for the DOJ Antitrust Division, Jonathan Kanter, recently said that “when we think about AI, we think about it as a tool . . . it’s really important that we understood that, so we’ve hired data scientists and are bringing in expertise to make sure we have the ability to understand that technology.”^[2]

As startup companies, venture capitalists, private equity funds, and other businesses compete to build teams and amass computational power necessary for generative AI applications, the FTC’s blog is a helpful reminder of areas where competition might be at risk.

For good measure, we also asked ChatGPT^[3] for its thoughts on antitrust concerns in the generative AI markets.

Question: What are the top three antitrust concerns in generative AI markets?

ChatGPT Answer: While antitrust concerns in generative AI markets can vary depending on specific circumstances, here are three prominent antitrust concerns that frequently arise:

1. Monopoly Power and Dominant Players: One of the primary antitrust concerns in generative AI markets is the potential for dominant players to acquire monopoly power. This can occur when a single company or a small group of companies control a substantial portion of the market, leading to reduced competition, limited innovation, and potential harm to consumer welfare. The concentration of power in the hands of a few entities may hinder market entry for smaller competitors and restrict consumer choice.
   
   
2. Exclusionary Practices: Antitrust concerns also arise when dominant players engage in exclusionary practices to maintain or strengthen their market position. These practices may include predatory pricing, bundling, or exclusive contracts that impede or deter competitors from entering or expanding in the market. By engaging in such practices, dominant firms can undermine competition and create barriers to entry, thereby limiting the emergence of new innovative players.
   
   
3. Control of Essential Inputs: Generative AI models rely on access to large and diverse datasets, computational resources, and other essential inputs. Antitrust concerns can arise when dominant players control or restrict access to these critical inputs, thereby hindering competitors’ ability to develop and improve their own generative AI technologies. By controlling access to essential inputs, dominant firms can impede competition and stifle innovation in the generative AI market.

Addressing these antitrust concerns requires regulatory scrutiny and enforcement to ensure fair competition, prevent the abuse of market power, and promote a level playing field for all participants in the generative AI market. By fostering competition and encouraging market entry, innovation can thrive, leading to more diverse and beneficial outcomes for consumers and society as a whole.

---