OpenAI, Anthropic and Google secretly joined forces to collaborate on AI safety

OpenAI, Anthropic, and Google have collaborated secretly to develop a framework for AI safety, according to Chris Lehane, the global head of policy at OpenAI. At a press conference in Washington, Lehane stated that his company, along with rivals Anthropic and Google's DeepMind research lab, has been working together to prioritize safety.

Anthropic's CEO, Dario Amodei, had previously called for cooperation to slow down the rapid development of frontier AI to avoid catastrophic risks. All major rivals, including OpenAI CEO Sam Altman, Elon Musk of SpaceXAI Corp., and Google's Demis Hassabis, supported Amodei's initiative. Despite concerns about potential antitrust violations, Lehane believes a waiver may not be necessary.

Legal experts warn that such coordination could be seen as "output restriction" under the Sherman Act. Critics argue that this collaboration could be an attempt by the top AI labs to maintain their dominance in frontier model development, potentially creating a cartel that could pose an existential threat to the U.S. in the AI race.

OpenAI is reportedly backing two bipartisan bills, the FRONTIER Act and another yet to be developed, which aim to create a federal framework for AI safety and give the Commerce Secretary the power to conduct safety audits before model release.