DARPA-NSF-CAISI AI Forge launches university RFI for national security AI
New $750K–$3M project ventures target interpretability, control, and adversarial robustness.
The DARPA-NSF-CAISI AI Forge program has entered its next phase, unveiled via a LessWrong post by Mike Vaiana and colleagues. The initiative released two key milestones: the 'Critical AI Challenges for National Security' report—developed with eight frontier AI companies and over fifteen Chief AI Officers from the Department of War and Intelligence Community—and a formal RFI (DARPA-SN-26-80) targeting U.S. universities. The program aims to bridge academic research with industry-scale compute and capabilities to solve pressing national security AI gaps. Universities can submit single, VP-authorized responses by June 22 for Project Ventures ranging from $750K to $3M or higher, spanning up to one year.
The three strategic thrusts are AI Interpretability (actionable understanding of models), AI Control (ensuring reliable performance under stress), and Adversarial Robustness (building secure AI for contested environments). The RFI responses will directly shape the upcoming solicitation for abstracts and the first AI Forge Pitch Day. Program Manager Matthew Marge emphasized that combining academic rigor with industry scale is key to achieving 'the impossible' in AI alignment and military applications.
- DARPA-SN-26-80 RFI targets U.S. universities for $750K–$3M+ Project Ventures up to one year.
- Three strategic thrusts: AI Interpretability, AI Control, and Adversarial Robustness for national security.
- Report co-developed with eight frontier AI companies and 15+ Chief AI Officers from DoW and IC.
Why It Matters
Bridges academic AI research with government and industry to solve critical national security challenges at scale.