Overview
- Reports published Friday say the three companies are finalizing plans for a privately run Standards Authority for Frontier AI, provisionally called SAFA, with a possible launch in late 2026 or early 2027.
- Organizers have approached former White House AI adviser Sriram Krishnan and ex‑OSTP director Arati Prabhakar to lead the body and have also discussed other high‑profile board candidates.
- Key design questions remain unresolved, including whether SAFA would perform in‑house pre‑deployment model assessments, how it would enforce rules, and what legal authority it would hold.
- Critics warn that if the largest vendors set accreditation and audit rules, compliance costs could rise and create entry barriers that squeeze out smaller and open‑source developers.
- The move follows stalled federal oversight talks and public safety calls from industry figures, and it could slow some product rollouts while giving enterprises clearer, voluntary benchmarks for model safety.