OpenAI, Anthropic and Google are working to create an AI standards body

Summary

Reports indicate that OpenAI, Anthropic, and Google are collaborating on an oversight body to test powerful AI systems before public release. The proposal was initially raised by Google DeepMind. The initiative is positioned as a response to growing security concerns and regulatory pressure in the sector.

Why it matters

If a pre-release testing gate for frontier models is implemented, it would represent the first industry-wide self-regulatory oversight point of its kind. This could also shape how governments approach AI oversight. The fact that three direct competitors are converging around a common standard mechanism signals a shift from unilateral security claims to a (theoretically) verifiable baseline.

Highlights

• The oversight body is still in the 'preparation phase' and has not yet become operational
• The proposal was raised by Google DeepMind; this raises questions about agenda-setting power
• The stated motivation is external: security concerns and regulatory demands; it is not a purely voluntary step

Critical take

It is essentially a two-sentence announcement that lacks any details regarding governance structure, enforcement authority, membership criteria, or what 'testing' concretely entails. This omission makes it impossible to assess whether the mechanism is a meaningful safety measure or merely a PR-friendly checkbox. The absence of independent or public sector voices in the described structure also leaves the question of who holds the final say unresolved.