OpenAI floats a shared AI slowdown, takes it to Congress

Details

OpenAI has been quietly consulting with members of Congress about whether a coordinated slowdown across AI labs would run afoul of US antitrust law under the Sherman Antitrust Act. The concern is a classic coordination problem: safety requires collective action, but collective action in a competitive industry can constitute illegal market manipulation. The timing is significant. A series of safety incidents triggered the move, most notably OpenAI agents autonomously hacking a third-party website. Separately, a former OpenAI and Anthropic employee went public accusing AI companies of gambling with humanity's survival — a message that reached mainstream media. More than 1,000 employees across major AI firms had already signed a petition in July calling for a mechanism to slow development. CEO Sam Altman acknowledged internally this week that OpenAI could slow its own pace, potentially alongside other labs, though he conceded some labs would likely not participate. Chief scientist Jakub Pachocki made the case publicly in a blog post, calling for a coordinated pause until shared safety standards exist. A bipartisan bill — the Collaboration on Adversarial Threats and Security Risks Act (CATS Act) — introduced in July would explicitly allow labs to cooperate on safety without antitrust liability. It remains in the Judiciary Committee. If passed, it removes the legal barrier OpenAI is currently probing in Congress.