OpenAI confirms safety talks with Anthropic and Google DeepMind

OpenAI confirmed it has been talking with Anthropic and Google about how the companies can work together on AI safety, a spokesperson told CNBC. The discussions have been underway since Google DeepMind chair Demis Hassabis floated a U.S.-led Standards Body in July.

The rivals compete hard on models and talent. They are also under growing pressure on safety and security. A joint posture would be unusual, which is why the confirmation matters.

What Hassabis proposed

Hassabis published an essay on X calling for a public-private partnership or self-regulatory organization with federal oversight, CNBC reported, comparing the idea to the Financial Industry Regulatory Authority. Screens for advanced models and coordinated slowdowns if risks rose were part of that framing in later coverage.

OpenAI’s confirmation tracks reporting that The Information had earlier described talks among the companies. Google and Anthropic did not immediately respond to CNBC’s request for comment.

Amodei’s slowdown call sits in the background

On Saturday, Anthropic CEO Dario Amodei urged labs to slow how quickly they improve their most advanced models. Hassabis and OpenAI CEO Sam Altman publicly backed the direction. Altman said pacing frontier models has been a primary topic inside OpenAI and that the company would share more soon.

Chris Lehane, OpenAI’s chief global affairs officer, wrote earlier this month that OpenAI wants to advance industry-led standards, with or without government support. He said voluntary efforts would complement, not replace, mandatory federal safeguards.

The AI Insider reported that Lehane told reporters coordination has run for several weeks and that the companies do not need a government waiver to continue safety work, even as some figures, including Altman in other comments, have flagged antitrust risk if collaboration looks like suppressing competition.

Lehane also said OpenAI backs a FRONTIER Act provision that would require frontier labs to allow independent verification of safety practices, according to that report. President Donald Trump has dismissed AI safety alarms and warned that slowing development would help China, coverage noted.

Whether a standards body actually forms, and whether it can bind competitors when incentives diverge, remains open. For now the news is narrower and still striking: OpenAI is saying out loud that it is talking safety with Anthropic and Google DeepMind.