Google, OpenAI and Anthropic Are Reportedly Working on an AI Safety Standards Body

Google, OpenAI and Anthropic are reportedly working on a voluntary safety standards body for the most advanced AI systems. It could launch in late 2026 or 2027. The consequential question is not yet which tests it will use, but who will have a say in writing them.

A proposed body, not a rulebook

The tentative name is Standards Authority for Frontier AI. Despite that name, the companies have not publicly announced its membership, powers or standards. The proposed organization would operate without government oversight, according to the account of the plan.

A safety standard could give developers a common way to assess a model before release. But a shared method is not the same as an enforceable decision: someone would still need to decide who conducts evaluations, what results require action and whether participants must comply. None of those powers has been publicly specified for this proposed body.

What would it add to the existing forum?

The three companies already participate in the Frontier Model Forum, founded in 2023 to advance safety research and best practices. The new organization's distinct role has not been explained. Will it develop testing methods, assess whether members meet them or simply publish guidance? Those are different jobs, and the proposal has not publicly assigned them.

That distinction matters when judging the plan. A forum for exchanging research and best practices need not have authority over its participants. A standards body might do more, but its name alone cannot establish what it would require of a developer whose model falls short.

Voluntary standards and mandatory rules

OpenAI has said it supports working with other frontier labs on voluntary standards, with or without government support. It also advocates mandatory U.S. safety requirements. Those positions are not necessarily in conflict: industry members could agree on shared measures while seeking legal obligations from lawmakers. The reported plan, however, does not itself create those obligations.

For now, the plan offers a possible venue for setting common expectations, not a public test of whether the companies will meet them. Its eventual membership, treatment of smaller developers and independence of any evaluators would help show whether it serves as a credible check or primarily reflects the priorities of its founders. Those design choices remain open.