Google, OpenAI and Anthropic AI Safety Group Takes Shape

Leo Schwartz and Stephanie Palazzolo: Google, OpenAI and Anthropic are advancing a self-regulatory AI safety standards body—tentative name Standards Authority for Frontier AI (SAFA)—aiming to launch by end-2026 or early 2027 without government oversight, per people familiar. Would support third-party pre-deployment testing, incident-reporting practices, operationalize prior voluntary safety commitments, and set auditor qualifications; debating whether SAFA itself tests models atop Commerce’s CAISI. Leadership shortlists (some not approached): CEO candidates Sriram Krishnan and Arati Prabhakar; chair shortlist Condoleezza Rice and David Friedberg; science advisers Beth Barnes (METR) and Paul Christiano (CAISI adviser / OpenAI foundation board). Public-private FINRA-style path (Hassabis essay; drafted EO) stalled after Meta/Nvidia skepticism; White House told labs to find industry consensus first. Overlap with Frontier Model Forum (Google/OpenAI/Anthropic/Microsoft, 2023) draws “performative” critique; Meta and many open-source builders likely outside initial “frontier” remit—critics fear competitive bar. Altman at UN Security Council: major AI decisions must go through democratic governments. Caveats: people-familiar; names tentative; candidates may not be active; launch timing aspirational.