OpenAI, Anthropic, Google DeepMind Hold Weeks-Long AI Safety Talks

On Tuesday, Chris Lehane, OpenAI’s Chief Global Affairs Officer, stated that the company has been in discussions with competitors Anthropic and Google DeepMind regarding AI safety issues for several weeks. He is engaging with U.S. lawmakers in Washington, focusing on catastrophic risks posed by frontier AI and corresponding governance measures. The collaboration was spurred by recent remarks from Dario Amodei, CEO of Anthropic. Amodei called on the industry to slow the expansion of frontier AI capabilities to buy time for building safety mechanisms. This initiative subsequently received support from Sam Altman, CEO of OpenAI; Demis Hassabis, head of Google DeepMind; and Elon Musk, CEO of SpaceX AI. Altman also noted that OpenAI is willing to bring in third-party evaluators to participate in safety oversight. Reports indicate that the three companies are also discussing establishing an industry body for AI standards. Because coordination among enterprises may raise antitrust concerns, Amodei had previously suggested limited government exemptions; however, Lehane argued that cooperation centered on security does not require such arrangements. OpenAI simultaneously supports provisions related to the FRONTIER Act proposed by bipartisan lawmakers. The bill would require major developers of frontier AI to hire licensed independent auditors to continuously assess corporate risk management, safety governance, and catastrophic risk mitigation measures.