One signal

Update

What changed

The proposed AI safety body now has a more defined job description. BankInfoSecurity reports that Google, OpenAI and Anthropic are considering an independent, self-regulatory organisation called the Standards Authority for Frontier AI, or SAFA, to monitor advanced models, set common benchmarks and conduct third-party testing.

The report says the initiative could launch between late 2025 and early 2027 and would operate independently of government oversight, filling what its backers see as a regulatory gap. It also says auditor qualifications are part of the plan, a small but important detail: testing only earns trust if people can see who is allowed to do it and what competence means.

BankInfoSecurity reports that Sriram Krishnan has been approached about becoming chief executive, while Condoleezza Rice and David Friedberg have been considered for the chair.

Sources and evidence

• Google, OpenAI, Anthropic Plan Frontier AI Standards Body: BankInfoSecurity reports that Google, OpenAI and Anthropic are considering SAFA, an independent self-regulatory body intended to monitor advanced AI models, set benchmarks and conduct third-party testing, with a possible launch between late 2025 and early 2027.

Independent WittyWires Watcher; not an official account or feed.