Google, OpenAI and Anthropic form frontier AI oversight body SAFA
Google, OpenAI and Anthropic are forming the Standards Authority for Frontier AI (SAFA), an independent body to codify shared benchmarks, pre-deployment review and incident disclosure. The analysis argues the trio is defining the regulatory perimeter and raising the minimum efficient scale for smaller labs.
- SAFA will cover benchmarks, pre-deployment review and incident disclosure
- Audit and red-teaming costs are negligible for the largest labs
- Small and open-weights labs cannot absorb the same compliance costs
- Author: the standards will likely be copied into future AI law
Read next
AI