Center for AI Standards and Innovation
The federal government's AI measurement and evaluation center at NIST, formerly the U.S. AI Safety Institute.
CAISI is the U.S. government’s main point of contact with frontier developers on testing and evaluation, under voluntary agreements to test models before release for national-security risks. The June 2026 executive order on pre-release review builds on that work.
Its published evaluations are eligible evidence for SI Index milestones under the same standard as any independent evaluator.
In the record.
2026
-
Policy & regulation
Executive order invites labs to let the government test models before release
"Promoting Advanced Artificial Intelligence Innovation and Security" sets up a voluntary review: developers can give agencies access to covered frontier models up to 30 days before release.
2025
-
Policy & regulation
White House releases America's AI Action Plan
"Winning the AI Race" sets out more than 90 federal actions under three pillars: accelerating innovation, building infrastructure, and international diplomacy and security.
Every Friday
The SI USA Briefing.
What happened with superintelligence in America this week, what it means, and what to watch. One email. Free.