OpenAI, Anthropic and Google float an AI standards body
CNN reports people familiar with private talks say Anthropic, Google, and OpenAI have discussed an AI industry standards body. OpenAI's chief scientist said more on external standards talks would come in the next months. No body has launched.
Anthropic, Google, and OpenAI have discussed creating an AI industry standards body, according to people familiar with the private talks.
CNN Business reported the conversations on 14 September 2026. The Information first reported them. The talks are unfinished. No body has launched, and representatives for Google, OpenAI, and Anthropic declined to comment or did not respond to CNN.
CNN says the talks preceded the past week's resignation drama and Anthropic CEO Dario Amodei's public pacing essay. The catalyst, according to those familiar with the situation, was a July essay by Google DeepMind founder Demis Hassabis. He proposed a US-led standards body modeled on the Financial Industry Regulatory Authority (FINRA), with public-private oversight, industry funding, and independent technical experts testing advanced models before deployment.
The conversations are ongoing, one person told CNN, with or without the Trump administration. Bloomberg had reported Treasury Secretary Scott Bessent was considering a FINRA-like AI agency that would report to the Securities and Exchange Commission. Meta CEO Mark Zuckerberg reportedly advised President Trump against the idea in a summer phone call, Politico reported, via CNN.
OpenAI chief scientist Jakub Pachocki told reporters the lab has been talking "to some external organizations about potential concrete standards" and would have more "in the next months." Sam Altman said late Sunday they look forward to collaborating across the industry on the best version of industry standards.
House Speaker Mike Johnson told CNN there is little consensus among the competing companies on what the guardrails should be. The White House still has an opaque voluntary pre-release review window of up to 30 days. Eligibility and process are not publicized.
The same labs are already writing their own safety paperwork in public, including Anthropic's enterprise frontier safeguards. A shared third-party eval window would be a different animal: one set of tests, not three vendor policies.
Some offers on this page may be paid placements or contain affiliate links.
Subscribe to Techpresso
Free daily newsletter, read in 5 minutes.
Subscribe free