Microsoft, Google and xAI have agreed to provide the US government with early access to their new artificial intelligence models to enable safety evaluations before they are brought to market. This agreement, coordinated by the Center for AI Standards and Innovation (CAISI) under the Department of Commerce, aims to analyze the capabilities of these systems and identify potential vulnerabilities, particularly regarding national security.

This initiative comes amid growing concerns over the rapid development of advanced AI models, which could be exploited to bolster cyberattacks. Recent progress, such as certain models specializing in cybersecurity, has heightened these concerns. CAISI, which is responsible for these evaluations, stated that it has already conducted over 40 tests, including on models not yet available to the public, with the cooperation of the companies involved.

This framework follows agreements reached in 2024 with other major industry players. At the same time, the Department of Defense is intensifying its integration of AI by signing partnerships with several tech firms to deploy these tools on classified networks, as part of a strategy to strengthen its capabilities and diversify its suppliers.