OpenAI, Anthropic and Google DeepMind hold weekly safety talks as Trump dismisses risks

Chris Lehane, OpenAI's head of global policy, confirmed to reporters in Washington on Tuesday that the company has been in ongoing discussions with Anthropic and Google DeepMind on artificial-intelligence safety for several weeks. Bloomberg first reported the talks. Lehane is in the capital to work with lawmakers on the catastrophic risks associated with frontier models.

The disclosure follows a Saturday essay by Anthropic chief executive Dario Amodei urging the industry to cooperate in slowing the pace of frontier-model development and avoiding existential risks. Leaders across the sector — OpenAI's Sam Altman, Google's Demis Hassabis and xAI's Elon Musk — voiced support for the call. Altman also said OpenAI would join Anthropic in embedding third-party evaluators inside their companies to monitor safety.

Lehane acknowledged at the briefing that the three firms have been talking for weeks, but said such coordination could expose them to antitrust suits if found to suppress competition. Amodei has proposed a narrow government exemption to allow safety coordination; Lehane argued the companies do not need one. In a Fortune interview last weekend, Altman hinted that private conversations with other industry leaders were already underway. The Information reported this week that the three are working to establish an industry standards body, a move Altman said would need to proceed without U.S. government backing.

Hassabis pushed for a formal watchdog as early as July, calling on the U.S. government to create a new standards body with authority to examine the world's most advanced models and coordinate industry-wide slowdowns if risks escalate. For now, President Trump rejects safety concerns as a "hoax" and opposes tighter regulation, arguing any slowdown would hand China an advantage in the race. His senior AI adviser, David Sacks, a veteran investor with personal stakes across much of the industry, reinforced that line, calling existential-risk fears exaggerated.

At the same Tuesday session, Lehane said OpenAI supports a provision in the FRONTIER Act that would require leading frontier labs to admit "independent verification organizations" into their facilities to verify that models are developed safely. If enacted, the measure would turn the voluntary third-party evaluator model into a binding legal requirement.