CAISI Secures Early AI Access from Google, Microsoft
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- CAISI announced on May 5, 2026 that it has secured agreements with five AI developers to give the U.S. government early access to their models before public launch.
- Google will provide early access to its frontier AI models under the new agreement, joining other firms in the government review process.
- Microsoft similarly agreed to share its upcoming AI systems for pre‑release security evaluation by U.S. authorities.
- xAI joined the pact, extending the early‑access arrangement to its AI models alongside the larger firms.
- OpenAI will also grant the government early access to its models, with Anthropic similarly participating in the review program.
- U.S. government will be able to evaluate the models for national‑security and safety risks before they are released publicly, as outlined by CAISI.
Why it matters: The agreement gives the U.S. government direct early‑access to evaluate frontier AI models for security and safety, allowing it to identify risks before public deployment; the participating firms must submit their models for review, potentially affecting release timelines and requiring additional compliance steps.
