Â鶹´«Ã½

The Research & Innovation Office (RIO) is sharing this internal call ahead of the anticipated solicitation release on October 23 to help faculty identify potential research directions and form teams in advance of the short submission timeline.Ìý

Below is a summary assembled by the Research & Innovation Office (RIO). This internal call may be updated as additional details become available in the full solicitation.

Program Summary

DARPA and the U.S. National Science Foundation (NSF) have jointly developed AI Forge to catalyze breakthroughs in AI for national security, working in close collaboration with the Center for AI Standards and Innovation (CAISI) at the National Institute of Standards and Technology.

AI Forge seeks disruptive scientific advances that make frontier AI1 systems interpretable, controllable, reliable, and secure in high-stakes, adversarial national security missions. This Cycle 2 report defines nine research and development (R&D) challenges for university-led Project Ventures across four complementary thrusts: AI interpretability, AI control, adversarial robustness, and AI assurance. Each challenge specifies a driving question, capability gap, scientific breakthrough sought, evidence of progress, and operational contribution. Cycle 2 emphasizes increasingly autonomous agents that operate over time with tools, persistent memory, delegated tasks, and changing conditions.

The four research thrusts address connected capability gaps.

Thrust 1 – AI Interpretability:ÌýEnabling Actionable Understanding and Auditing

• Research Challenges:Ìý
  • 1.1 Predict Emerging Failure
  • 1.2 Establish Causal Accountability
• Capability question: Can we detect emerging failures early enough to act and establish what caused them and how to correct them?

Thrust 2 – AI Control: Maintaining Human Authority Over AI

• Research challenges:Ìý
  • 2.1 Enforce Authorized Bounds
  • 2.2 Verify Steerability as Objectives and Conditions Change
  • 2.3 Control Autonomous R&D and Experimentation
• Capability question: Can we maintain effective authority over agents as mission objectives change and automated AI capabilities grow?

Thrust 3 – Adversarial Robustness: Securing Resilient AI for Contested Environments

• Research challenges:Ìý
  • 3.1 Preserve Mission Integrity Against Adaptive Adversaries
  • 3.2 Sustain Secure and Available AI Through Compromise and Disruption
• Capability question: Can we preserve mission capability as adversaries adapt, and restore secure operations after compromise or disruption?

Thrust 4 – AI Assurance: Evaluation and Verification for Mission Confidence

• Research challenges:Ìý
  • 4.1 Verify Scientific Outputs of AI Systems that Exceed Human Expertise
  • 4.2 Maintain Defensible Mission Assurance Over Time
• Capability question: Can we verify difficult AI results and justify continued trust in a system for a particular mission as conditions change?

Deadlines

• CU Internal Deadline: 8:00am MT October 29, 2026
• NSF-DARPA Abstract Deadline: November 6, 2026
• NSF-DARPA Pitch Deadline (by invitation): December 8, 2026

Internal Application Requirements (all in PDF format)

• Project Summary (2 pages maximum): Please specify the Thrust area and associated research challenges and capability question. Teams should define a consequential failure, the capability needed to address it, and the evidence that would demonstrate a scientific breakthrough.
• Curriculum Vitae of the Lead PIÌý
• Internal Teaming Abstract (optional): Please provide a brief project abstract (3-4 sentences) for teaming purposes. By submitting this abstract, you give RIO permission to share your name and abstract with potential Â鶹´«Ã½ teams as part of the internal competition process, including if your proposal is not selected.

To access the online application, visit:

Eligibility

AI Forge aims to bring together universities, frontier AI companies, and government mission stakeholders through a joint effort between DARPA, the National Science Foundation (NSF), and the National Institute of Standards and Technology’s Center for Advancing Innovation and Standards for Super Intelligence (CAISSI).Ìý

Limited Submission Guidelines

Institutions are allowed to submit one abstract.ÌýÌý