OpenAI, Anthropic and Google have been negotiating a common security framework for weeks

OpenAI's head of public affairs confirms discussions between the three laboratories: a common standards body and third-party evaluators based inside the companies.

Chris Lehane, Global Public Affairs Director of OpenAI, confirmed to the press on September 15, 2026 that his company has been working for several weeks with its competitors Anthropic and Google DeepMind on common security measures. On the agenda of exchanges: the creation of a sector-specific standardization body, and the integration of independent evaluators into laboratories.

These discussions take on a special focus after the forum published on the previous Saturday by Dario Amodei, patron of Anthropic, who called on industry to collectively slow the development of border models to avoid catastrophic risks. His proposal included a governmental derogation allowing such coordination without falling under competition law. Sam Altman, Demis Hassabis and Elon Musk each provided public support for the process, with Altman indicating that OpenAI would join Anthropic by installing third-party security reviewers.

The political welcome is cold. Donald Trump described security concerns as "canal," while his IA advisor David Sacks argues that an American slowdown would benefit China mechanically.