Defend, adopt, and advance frontier AI.

The Frontier Security Institute, at the Center for AI Safety, provides independent, nonpartisan policy studies and analysis that connect America's defense and intelligence leaders with the AI industry to meet U.S. economic and national security needs.

Our thesis

What we do

Three areas of work.
One purpose.

We work in three areas, each built to support the National Security Enterprise: the Pentagon, the Intelligence Community, federal agencies, Congress, and our allies. Each brings together people from government, industry, and academia to work a real national security problem. Together, they track the technology through its life in national security: defending it, adopting it, and advancing American interests with it.

See our work →
Defend

Model Security

Protecting U.S. frontier AI from theft, disruption, and adversarial attack. We work directly with the frontier labs to close the gap between the threats they face and the authorities the government already has to act.

In the near term, we work across technical, legal, and operational means to close the information gaps that defenders face. Looking further out, we examine the technical and governance structures that can keep American AI ahead for the long term.

Adopt

AI Adoption & Evaluation

Making sure AI is actually fit for the government's purpose. National security has needs that commercial benchmarks don't measure: reliability where failure isn't an option, and security requirements all its own.

We develop clear criteria for testing, evaluating, and fielding AI-enabled systems, so that developers and government users understand each other's needs, acquisition and operational risk goes down, and AI delivers from the tactical level to the strategic.

Advance

U.S. Capability for Global AI Governance

Advancing U.S. interests as AI capability accelerates. Working with national security veterans who have negotiated directly with America's adversaries, we study how the AI race reshapes American power.

We lay out how the United States can advance global governance of AI: managing competition, lowering the risk of miscalculation, and setting durable rules rather than trading away leverage to get them.

The translation layer

The work in between.

Running through all three programs is one job: translation. The people building frontier AI and the people responsible for the nation's security are converging on the same problem from two different worlds, with different languages, incentives, and speed. We convene them at the same table and translate between them, so that analysis reaches the people who can act on it.

Isaac “Ike” Harris, Executive Director, Frontier Security Institute

Why us

We know how these decisions get made.

AI adoption and U.S. strategic advantage are mutually reinforcing, not competing.

Our team has worked at the senior levels of the uniformed services, the Intelligence Community, the Department of Commerce, and Congress. That experience shapes the questions we ask and the rigor we bring to answering them.

Meet the team

Work with us.

If you build, field, govern, or oversee frontier AI in a national security context, that’s what we’re here for.

Contact Us