10 Ways AI Labs Can Limit Antitrust Risks of Safety Pacts

Attorneys Jeffrey Seul, David Kully and Marissa Serafino wrote an article for Law360 exploring the antitrust implications of agreements to limit artificial intelligence (AI) development and listing 10 strategies to combat the resulting compliance risks. With more and more industry leaders calling on companies to create checkpoints and slow development, particularly following the OpenAI-Hugging Face incident, horizontal coordination among competitors may yield safety benefits but also invites scrutiny under Section 1 of the Sherman Antitrust Act. As the authors explain, agreements to restrict output for no purpose other than restricting output are considered illegal per se, whereas legitimate pacts that allow participants to achieve together what they could not alone may pass muster. With this in mind, their article goes on to describe ways companies can structure AI pacing agreements to maximize safety while minimizing legal risks. Measures such as making participation voluntary and easily withdrawn, confining data sharing to safety-related matters, employing third-party evaluators, documenting procompetitive justification and allocating liability can all go a long way toward managing antitrust exposure and fending off potential enforcement as the industry works toward reform and enhanced protection.

READ: 10 Ways AI Labs Can Limit Antitrust Risks of Safety Pacts