Major AI firms join government-led safety reviews
Google DeepMind, Microsoft, and Elon Musk's xAI have agreed to let the Commerce Department's Center for AI Standards and Innovation (CAISI) review their new frontier AI models prior to public release. CAISI will perform pre-deployment evaluations and targeted research to better assess capabilities and potential risks, extending a safety pathway the center has been building with other industry players.
CAISI began evaluating models from OpenAI and Anthropic in 2024 and has completed roughly 40 reviews so far. Adding these three large developers broadens the program's reach and creates a more consistent, shared approach to safety testing across the companies building the most advanced models.
Benefits of the collaboration include:
- Stronger pre-release safety checks to reduce harms at deployment;
- Targeted research that improves understanding of frontier capabilities and weaknesses;
- Increased transparency and coordination between industry and government, which can boost public trust and inform sensible policy.
By scaling CAISI's review work to include more major developers, the industry takes a practical step toward safer rollouts of powerful models. The agreement signals a constructive model of cooperation where government testing and company innovation work together to deliver safer AI to the public.