Anthropic, Google and OpenAI have been talking among themselves about setting up an AI industry standards body, according to CNN, with The Information first reporting the talks. The trigger was Demis Hassabis’s July essay proposing a US-led body modeled on FINRA that would test advanced models before they ship, sit under government oversight, and be paid for by the industry. One person said the conversations began before an Anthropic researcher resigned last week saying the leading labs weren’t behaving responsibly, and that they’ll continue whether or not the Trump administration takes part. Nothing binds anyone today beyond a voluntary White House review companies can use in the month before a release, whose eligibility rules have never been published.
