OpenAI Wants to Know if an AI Industry Slowdown Would Even Be Legal
wired ·
OpenAI has asked members of Congress in recent weeks for clear guidance on whether orchestrating an industry-wide slowdown in frontier AI development would be legal, according to people close to the company. Those people say substantive safety coordination between AI labs may risk violating antitrust law, which poses a major obstacle to getting large tech companies to join the effort. OpenAI chief scientist Jakub Pachocki argued in a weekend blog post that coordinating to slow future development is the best path forward and that voluntary slowdowns may become common until shared safety bars are set. Some legal scholars say such coordination could conflict with US antitrust laws; Nicholas Felstead wrote in March that a coordinated pause may amount to restricting output and potentially violate the Sherman Antitrust Act, though it would depend on the agreement’s precise details. In July, a bipartisan, bicameral group introduced the Collaboration on Adversarial Threats and Security Risks Act, which would explicitly allow AI labs to coordinate on security and safety without antitrust risk, but the House version has not been taken up. Caleb Knapp of the AI Policy Network said the bill would create legal channels for labs to address safety incidents and that Congress has growing appetite to act, though legislation may wait until after the midterm elections. Some AI leaders argue antitrust concerns are convenient cover, pointing instead to fierce business competition, national security views about staying ahead of China, and differing opinions on how to develop safe AI. OpenAI cofounder John Schulman, now chief scientist at Thinking Machines, wrote on X that OpenAI and Anthropic should stop feuding and jointly develop a pacing proposal, saying antitrust does not prohibit jointly developing a proposal. The article also notes rising alarm over the AI race, including a public warning from former Anthropic and OpenAI researcher Jacob Coxon and security incidents such as OpenAI’s agents hacking Hugging Face.
wired