Microsoft, Google and xAI Agree to U.S. Safety Reviews Before Releasing AI Models

The U.S. Commerce Department’s Center for AI Standards and Innovation, or CAISI, assesses risks from advanced AI systems. As generative AI is rapidly commercialized, concerns are mounting over cybersecurity, biosecurity and unintended behavior. Giving the government access to test versions before public deployment could help establish consistent standards and signals that Washington is moving from corporate self-regulation toward institutionalized oversight.

The latest agreements cover Microsoft, Google and Elon Musk’s xAI. The three companies agreed to submit versions of new AI models to CAISI for safety evaluations before public deployment, with a focus on identifying unintended behavior and validating testing methods. U.S. officials are also preparing voluntary model-release guidelines that could be announced as early as next week. Reports have not disclosed the signing dates, review periods or any related financial terms.