OpenAI, Anthropic and Google Explore an Industry AI Standards Body

NEW YORK - OpenAI said on Sept. 15 that it has been talking with Anthropic and Google about how the companies can cooperate on artificial intelligence safety, including an idea for an industry standards body modeled on U.S. financial self-regulation.

A company spokesperson told CNBC that the conversations have continued since Demis Hassabis, chair of Google DeepMind, published a July proposal for a U.S.-led “Standards Body.” Hassabis sketched a public-private organization with federal oversight, funded largely by industry, that would test advanced models before wide release. He pointed to the Financial Industry Regulatory Authority, which oversees brokers, as a template.

Photo: Alain Herzog / Wikimedia Commons

Photo: Anthropic

The Information first reported that Anthropic, Google, and OpenAI had discussed creating such a body. CNN later said people familiar with the talks described the same exploratory track. Google and Anthropic did not immediately comment to CNBC.

No charter, membership rules, or testing mandate has been announced. OpenAI’s chief global affairs officer, Chris Lehane, wrote earlier this month that the company wants voluntary industry-led standards “with or without government support,” and that any such effort would complement, not replace, mandatory federal safeguards. Bloomberg reported that Lehane said engagement with Anthropic and Google had been under way for several weeks and that OpenAI does not see a need for an antitrust waiver to coordinate on safety.

The talks land in a week of heightened pressure on frontier labs. On Sept. 12, Anthropic chief executive Dario Amodei urged companies to slow how quickly they raise the capability of their most advanced models. OpenAI chief executive Sam Altman, Elon Musk, and Hassabis publicly backed the direction even as details remain unsettled. Separately, President Donald Trump has dismissed some AI risk warnings as overblown, a stance that keeps Washington’s appetite for new rules uncertain.

For now the story is process, not product. Three of the largest model makers acknowledge they are exploring shared standards while they still compete for talent, customers, and chips. Whether that becomes a FINRA-style tester, a lighter voluntary code, or another stalled industry forum will depend on what they publish next, and on whether governments treat self-regulation as enough.