Microsoft, Alphabet-owned Google, and Elon Musk's xAI have agreed to provide the US government early access to new artificial intelligence models before public release, allowing for national security risk assessments.
The Center for AI Standards and Innovation (CAISI) at the Department of Commerce announced Tuesday that the agreement enables evaluation of models before deployment, including research into their capabilities and security risks.
This follows the emergence of advanced AI systems like Anthropic's Mythos, which raised global concerns among US officials and corporate leaders about their potential to empower hackers.
OpenAI's ChatGPT-maker recently unveiled GPT-5.4-Cyber, a variant fine-tuned for defensive cybersecurity work, after Anthropic's announcement of Mythos. Anthropic is currently in a dispute with the Pentagon over military use of its AI tools.
"Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," said CAISI Director Chris Fall.
The initiative builds on 2024 agreements with OpenAI and Anthropic, established under the Biden administration. CAISI has completed over 40 evaluations, including on cutting-edge models not yet public. Developers often provide stripped-back versions without safety guardrails for the center to probe national security risks.
Last week, the Pentagon reached agreements with seven AI companies to deploy advanced capabilities on classified networks, aiming to broaden AI providers across the military.