Does AI need an antitrust exemption so it doesn’t kill everyone????
Jonathan Kanter, former head of the DOJ Antitrust Division and current law and technology-policy professor, told Decoder that AI development needs clear legal "lines on the road" rather than an antitrust exemption allowing industry coordination. He argued companies already can collaborate on cybersecurity measures without special immunity, and that product-liability rules and government regulation should hold firms accountable for unsafe AI.

Why It Matters
The debate over whether frontier AI firms should get an antitrust exemption intersects urgent questions about safety, competition, and national policy toward China — and could reshape how fast and under what constraints advanced AI systems are developed. Kanter’s views draw on recent high-profile antitrust actions and ongoing regulatory uncertainty, making his take relevant to lawmakers, industry, and researchers grappling with governance choices.
Key Facts
- Interviewee: Jonathan Kanter, former DOJ Antitrust Division chief; professor at WashU and Carnegie Mellon
- Source program: Decoder (The Verge)
- Antitrust cases mentioned: Kanter led DOJ cases against Google, Ticketmaster, and Apple (Apple case remains pending)
- Companies named in discussion: OpenAI, Anthropic, Google DeepMind, Meta, Nvidia, Hugging Face
- Proposals discussed: Calls from some industry leaders for slowing AI development and requesting antitrust exemptions to coordinate on safety
Jonathan Kanter said the current AI moment resembles inventing cars without traffic rules: powerful technology is being deployed without clear legal constraints, creating risks that require both company responsibility and government action. He rejected the idea that seeking regulation should automatically become a cartel or justify an antitrust exemption, saying firms can and should build safe systems on their own and already may engage in legitimate information-sharing on security without special immunity. Kanter outlined two interpretations of companies’ calls for coordinated slowdowns: a generous reading that founders genuinely fear runaway harms, and a cynical one that firms want regulation to ease competitive pressure and shore up valuations ahead of IPOs. Neither scenario, he said, warrants an antitrust exemption. Instead, he argued that existing antitrust law allows for collaboration on safety (for example, a threat-clearinghouse) and that liability frameworks should attach to companies whose products cause harm. On liability, Kanter pointed to the recent products-liability litigation against Meta over teen harms as a precedent showing courts can hold platform owners accountable, though such cases can take years. He urged Congress to clarify legal consequences for dangerous AI — enumerating societal values like mental health, competition, authenticity of information, and intellectual property — and to codify those "lines on the road." Absent timely federal action, Kanter said states and existing liability doctrines will play roles in shaping incentives for safer products. Kanter pushed back against arguments that the United States must tolerate domestic monopolies to compete with China, calling that approach counterproductive. He said the better path is vigorous domestic competition combined with rules that protect users and public interests. Finally, reflecting on his tenure leading a more aggressive antitrust enforcement agenda, he said the movement toward stricter scrutiny of dominant tech firms has shifted the debate even if many enforcement remedies remain incomplete, and urged policies that protect the next inflection point for competition rather than freezing market power in place.
Keep Reading

East Asia: How AI is changing the way we bid farewell to loved ones
Doomsday AI: panic, regulation and the China fear

The AI regulation smackdown isn’t over
