Three frontier labs discussed writing their own rulebook

The Information reported on Sunday, September 13, 2026 that Anthropic, OpenAI and Google have held working-group meetings since July about creating an industry-led standards body for frontier AI models. PYMNTS, summarising that report the next day, said the discussions centre on technical testing and pre-release auditing. What is described is a series of meetings, not an organisation: no charter, membership rule or start date has been published.

The reported catalyst is a July essay by Google DeepMind’s Demis Hassabis proposing a self-regulatory body modelled on FINRA, the self-regulator for the US securities industry. PYMNTS reports that OpenAI’s Sam Altman has signalled support for a testing and auditing organisation while preferring that the labs set the standards themselves, without federal backing. The participants are reported to disagree on the body’s shape.

Timing shapes how the report reads. It landed the day after Anthropic’s Dario Amodei published an essay arguing that labs should deliberately slow capability gains, a position Altman, Hassabis and Elon Musk endorsed publicly over that weekend. A private standards discussion and a public pacing argument are separate things. The reporting places them in sequence; it does not establish them as one agreed programme.

What the reporting does not settle is most of it: whether any body will be formed, which companies would join, whether labs outside the three would be in scope, what an audit would actually test, who would see the results, and what would happen when a member disagreed with a finding. None of the three companies confirmed the discussions to reporters at the time.