AI Forge fast-tracks breakthroughs for national security

DARPA and the National Science Foundation have launched AI Forge, a joint research and development program designed to steer cutting-edge artificial intelligence toward national security needs that fall outside mainstream commercial priorities. The effort, developed in collaboration with the National Institute of Standards and Technology’s Center for AI Standards and Innovation, is aligned with America’s AI Action Plan and aims to make AI more reliable in high-stakes environments, more understandable to operators, and more secure against adversaries. It also seeks to strengthen the pipeline of talent and ideas among universities, frontier AI companies, and government.

To shape its agenda, the program gathered representatives from leading AI firms, chief AI officers from more than 15 Department of War and Intelligence Community agencies, and other federal stakeholders. Their input underpins the newly released AI Forge Critical AI Challenges for National Security report, which outlines 15 research challenges across three thrust areas for university-led teams:

– AI interpretability: advancing beyond routine explanations toward operational understanding of model behavior, decisions, and impacts.
– AI control: developing tools that give strong, verifiable evidence of bounded, auditable, and reliable behavior today, while laying groundwork to preserve meaningful human control of more capable systems.
– Adversarial robustness: establishing scientific foundations for resilient-by-design AI that sustains integrity and intended performance under deliberate attack.

Program leaders say pre-competitive research in these areas could accelerate adoption of AI across industry and federal agencies. Given the rapid pace of technical progress, the list of challenges will be revisited every six months.

AI Forge is inviting universities to submit statements of capability in response to a Request for Information to help build a repository of U.S. institutions interested in advancing next-generation AI for national security. Responses are due by June 22.

Looking ahead, the program plans to stand up a nonprofit-administered forum this summer to fund, guide, and manage fast-paced, university-led projects. The forum is intended to pair academic expertise with frontier-scale compute, models, and industry know-how to tackle mission-driven problems identified by national security AI leaders across the Department of War and the Intelligence Community.

“We’re taking a unified approach to create breakthroughs in AI for national security,” said Matthew Marge, DARPA program manager for AI Forge. “The frontier AI companies build and commercialize massive, high-capability models and compute. Universities are engines of deep, foundational research and, importantly, they cultivate our nation’s future talent. At the intersection is high-risk, high-reward research that requires both massive scale and deep, mission-driven work – something that is difficult to pursue in either environment alone. With AI Forge, we’re looking to build an ambitious new ecosystem that bridges this gap.”

“NSF is excited to partner with DARPA, working alongside CAISI, on this groundbreaking effort to catalyze AI innovation for national security” said Erwin Gianchandani, NSF Assistant Director for Technology, Innovation and Partnerships. “By linking the rapid advances of frontier AI companies, the research and talent at universities, and the use cases surfaced by the intelligence community, AI Forge will propel advancements in AI capabilities for the benefit of U.S. national security and, ultimately, all Americans.”

Additional details on the forum and initial projects are expected in the coming weeks.