OpenAI Seeks Legal Clarity on Industry-Wide AI Slowdown

OpenAI has recently sought guidance from members of Congress regarding the legality of orchestrating an industry-wide slowdown in frontier AI development, as Wired first reported.

Jakub Pachocki, OpenAI's chief scientist, recently argued in a blog post that coordinating to slow future development is essential for the safety of self-improving AI systems. Pachocki suggested that voluntary slowdowns should become common until shared safety bars are established. However, legal experts warn that such coordination could conflict with U.S. antitrust laws. In a March article, Nicholas Felstead, assistant director of the Australian Competition and Consumer Commission, suggested that the Sherman Antitrust Act could be violated if a coordinated pause was interpreted as a restriction on output.

In response to these concerns, a bipartisan, bicameral group of lawmakers introduced the “Collaboration on Adversarial Threats and Security Risks Act” in July. According to Wired, this bill would explicitly allow AI labs to collaborate on safety and security without risking antitrust violations. Caleb Knapp, director of government affairs at the nonprofit AI Policy Network, stated the bill would create legal channels for labs to address security incidents, though he noted that enactment may wait until after the midterm elections.

Some industry figures, however, suggest antitrust concerns are a pretext. John Schulman, an OpenAI cofounder and current chief scientist at Thinking Machines, argued on X that antitrust laws do not prevent companies from jointly developing a proposal. Other factors contributing to the lack of coordination include fierce market competition, national security concerns regarding China, and differing philosophies on AI safety. These tensions emerge as security failures, such as OpenAI agents hacking Hugging Face, increase pressure from lawmakers for regulation.