Google, OpenAI and Anthropic Are Building Industry Standards Body SAFA

Per The Information reporters Stephanie Palazzolo and Leo Schwartz, Google, OpenAI and Anthropic are jointly building a voluntary frontier AI safety standards body named the Standards Authority for Frontier AI (SAFA), targeted to launch in late 2026 or early 2027. It could self-test frontier models before release rather than rely on government oversight. Working-group members have approached several prominent figures about the CEO role, including former White House AI advisor Krishnan; no pick has been finalized. The story remains a single-source report with no official confirmation from any of the three companies.

Confirmed

• Per The Information, the body is named the Standards Authority for Frontier AI (SAFA); some reposts render it as SAFFA, but the original reporting's spelling stands.
• It is a voluntary safety standards body jointly built by Google, OpenAI and Anthropic.
• Its functions include third-party model testing and incident reporting; per a Reddit repost, it may conduct pre-release self-testing without government oversight.
• Launch is targeted for late 2026 or early 2027.
• Per details relayed by @pstAsiatech, the group has approached multiple notable figures for CEO, including former White House AI advisor Krishnan, with no final choice.

Unconfirmed

• The report stems from a single outlet, The Information; none of the three companies has officially confirmed it.
• The final CEO pick, governance structure and membership scope remain unclear, as does whether Krishnan would accept.

Why it matters

• The three leading AI labs bypassing government-led regulation to set up their own safety standards body—and potentially self-testing models pre-release—could shape the future of frontier-model self-regulation and reshape dialogue with regulators. The presence of a former White House AI policy advisor among CEO candidates suggests personnel ties to US policy circles.
• @TansuYegen flagged the core tension: the rulemakers are the very people building the technology; self-regulation may be faster than waiting for government to understand frontier AI, but its independence is questionable—the main external concern about this model.

2026-09-24 ~ 2026-09-25 · 11 related posts

• Anthropic, OpenAI, Google building voluntary 'Standards Authority for Frontier AI' by 2027 — Hesamation · 2026-09-24
• [source] Google, OpenAI and Anthropic's new standards body named Standards Authority for Frontier AI — steph_palazzolo · 2026-09-24
• OpenAI, Google and Anthropic reportedly building SAFA, a frontier AI safety standards body — steph_palazzolo · 2026-09-25
• Google, OpenAI and Anthropic reportedly forming their own frontier-AI safety authority — 141_1337 · 2026-09-25
• [source] Google, OpenAI and Anthropic plan a self-run AI safety standards body without government oversight — pstAsiatech · 2026-09-25
• Frontier AI standards body to be named SAFFA; Sriram Krishnan approached as CEO — pstAsiatech · 2026-09-25
• OpenAI, Google and Anthropic reportedly building a shared AI safety standards body — TansuYegen · 2026-09-25

4 near-duplicate retellings: Hesamation · steph_palazzolo · TorturedPoet30 · emmanuelvivier