A National Center for Advanced AI Reliability and Security

This is a linkpost for a policy memo published by the Federation of American Scientists, which proposes scaling up a significantly enhanced “CAISI+” within the Department of Commerce.

CAISI+ would require dedicated high-security compute facilities, specialized talent, and an estimated annual operating budget of $67-155 million, with a setup cost of $155-275 million. CAISI+ would have expanded capacity for conducting advanced model evaluations for catastrophic risks, provide direct emergency assessments to the President and National Security Council (NSC), and drive critical AI reliability and security research, ensuring America is prepared to lead on AI and safeguard its national interests. The memo makes the following recommendations:

1. Establish CAISI+ to lead national AI safety and coordinate crisis response. Core functions should include advanced model evaluation, emergency assessment and response, and driving foundational research.

2. Equip CAISI+ with elite American talent and sustained funding, via special hiring authorities, competitive compensation, and appropriations to ensure a world-class federal core of AI expertise.

3. Equip CAISI+ with essential secure compute infrastructure. CAISI+ will require secure compute for evaluations involving proprietary models and national security data. We suggest a hybrid strategy featuring a modestly sized but highly secure SL-5 compute cluster for frontier model hosting and  analysis, and formal DOE/DoD SCIF and compute access for TS/SCI workloads that Commerce cannot originate.

4. Explore granting critical authorities. The executive branch should analyze existing federal powers for gaps in AI threat response capabilities and report to Congress where new authorities may be necessary.

5. Initiate activities (0-6 Months) to build CAISI+:

   1. Prioritize CAISI funding within existing appropriations, including specific allocation floors

   2. Grant CAISI a FOIA exemption for sensitive commercial information

   3. Leverage NIST's DOE lab relationships for interim classified computing access

   4. Initiate security research and procurement for the SL-5 compute facility

   5. Work with OMB and Commerce for initial funding (e.g., through reprogramming)

   6. Build on CAISI’s voluntary agreements to develop emergency model access protocols

   7. Begin OSTP-led analysis of federal authority gaps to manage AI threats

→ Click here to read the memo on the Federation of American Scientists’ website.