Google, Microsoft and xAI agree to US government AI testing programme
Google, Microsoft and xAI have agreed to submit their AI models for evaluation by the US Department of Commerce before public release. The Center for AI Standards and Innovation (CAISI) will assess the systems for cybersecurity, biosecurity and chemical weapons risks.
The evaluations mark a shift in the Trump administration's approach to AI oversight. In March, President Trump released an AI National Policy Framework emphasizing the removal of barriers to innovation and faster deployment across sectors. The framework opposed creating new federal regulatory bodies, instead directing existing agencies and domain experts to examine models.
CAISI Director Chris Fall said in a statement that "independent, rigorous measurement science is essential to understanding frontier AI and its national security implications."
What the companies are submitting
Microsoft said the evaluations will help it identify risks like AI-powered cyber attacks for its Copilot model.
OpenAI provided ChatGPT5.5 ahead of its public release this week for national security testing, according to Chris Lehane, the company's chief global affairs officer. OpenAI is also working with CAISI on GPT-5.5-Cyber, a model designed to strengthen cyber defence capabilities and currently available only to a limited group of users.
Anthropic and OpenAI signed initial agreements for these evaluations in 2024 under former President Joe Biden. CAISI said existing agreements have been "renegotiated," though it did not specify what changed.
CAISI's track record
CAISI has already conducted 40 evaluations of other models, including some unreleased state-of-the-art systems, but did not identify which ones.
For those working in government, understanding how AI systems are evaluated before deployment is becoming central to policy decisions. Learn more about AI for Government and how Generative AI and LLM systems are being assessed for risk.
Your membership also unlocks:
