Our Mission
Frontier AI and the National Security Mission.
For the Pentagon, the Intelligence Community, government offices, Congress, and our allies: independent, nonpartisan analysis of what frontier AI means for the mission.
How we work with the National Security Enterprise.
Government is adopting frontier AI as fast as it can, but it doesn’t build these systems. The choices that decide whether a system is secure and reliable get made during development, before government reaches testing and approval. The Frontier Security Institute helps the National Security Enterprise see what’s coming and get its requirements in early, independently and without a commercial or political stake in the result.
Briefings and written analysis
Off-the-record briefings for program offices, committee staff, and mission owners, timed to the decisions they inform, and short-form memoranda on the technical questions where an outside read matters.
Test and evaluation standards
What a system must be measured against for a specific national security use, built to fit the acquisition and testing processes already in place.
Wargames and scenario work
Structured exercises that stress-test how institutions would respond to AI-driven contingencies before they occur.
Testimony and oversight support
Nonpartisan analysis for the committees shaping the next decade of AI policy.
Classified work
Our staff hold security clearances, so the work can continue in a classified setting when it needs to.
The substance behind this work lives in our three programs:
Bring us into the room.
If the decision needs an independent voice fluent in both the lab and the field, that is our work.