DARPA AI: Programs, Funding Thrusts & Guide

DARPA AI represents the portfolio of advanced artificial intelligence research and development programs managed by the Defense Advanced Research Projects Agency (DARPA). Operating at the technological forefront of artificial intelligence, DARPA funds high-risk, high-payoff research to prevent technological surprise and secure national defense capabilities. Rather than funding routine commercial software, DARPA AI programs advance fundamental methodologies in machine learning, neuro-symbolic reasoning, and autonomous system verification.

What it is and who runs it

DARPA AI is administered primarily through the Information Innovation Office (detailed in our guide on DARPA I2O funding and thrusts), alongside collaborative efforts with the Defense Sciences Office (DSO) and Strategic Technology Office (STO). Program Managers (PMs)—expert scientists and engineers on rotating three- to five-year assignments—conceive, solicit, and direct individual research programs.

As described on the official DARPA program page on 2026-11-07, the agency coordinates its artificial intelligence portfolio through the landmark AI Next Campaign. Announced in September 2018 as a multi-year investment exceeding $2 billion across dozens of programs, AI Next frames machine learning development across three waves:

First Wave (Handcrafted Knowledge): Expert systems and rule-based logic capable of reasoning over narrow domains, but unable to learn or handle uncertainty.
Second Wave (Statistical Learning): Deep neural networks that achieve high classification accuracy on massive labeled datasets, but remain fragile and vulnerable to deception.
Third Wave (Contextual Adaptation): Advanced systems capable of acquiring explanatory models, reasoning with common sense, and functioning as collaborative partners with human operators.

Eligibility (summary, with the official source)

As specified in DARPA Broad Agency Announcements (BAAs) and official program announcements on darpa.mil, eligibility for DARPA AI funding is broad and competitive:

Higher Education Institutions: Accredited universities and degree-granting colleges (including early-career faculty applying for the DARPA Young Faculty Award).
Commercial Industry: Small businesses, defense contractors, and specialized artificial intelligence software startups.
Non-Profit and Independent Institutes: Federally Funded Research and Development Centers (FFRDCs) and university-affiliated research centers (UARCs), subject to organizational conflict-of-interest rules.
Foreign Entities: Foreign organizations and universities may participate when research is unclassified, complies with federal export control laws (ITAR/EAR), and satisfies foreign influence vetting.

What it funds or requires

DARPA AI programs fund basic research (Budget Activity 6.1), applied research (6.2), and advanced technology development (6.3). Project budgets range from $500,000 exploratory studies to multi-million-dollar multi-year performer contracts. Research thrusts span five primary capability areas:

Robust AI: Developing analytical and empirical methods to eliminate failure modes in computer vision, robotics, and reinforcement learning in unpredictable edge environments.
Adversarial AI: Creating mathematical defenses against input perturbation attacks, data poisoning during training phases, model inversion, and cyber exploitation.
High-Performance AI Hardware: Researching analog processing achieving 1000x speedup and 1000x power efficiency improvements over conventional digital chips, and algorithms reducing labeled data needs.
Explainable AI and Common Sense: Building systems that articulate the causal rationale behind their recommendations, moving beyond opaque “black box” statistical predictions.
Process Automation: Applying verified AI models to streamline core defense workflows, such as security clearance vetting and automated software accreditation.

Award mechanisms include standard procurement contracts under the Federal Acquisition Regulation (FAR), grants, cooperative agreements, and Other Transactions (OTs). Other Transactions are frequently utilized for non-traditional defense contractors and universities to enable flexible intellectual property arrangements.

Application or compliance steps

Applying for DARPA AI funding follows a structured procurement lifecycle:

Monitor Solicitations: Program Managers publish solicitations as Broad Agency Announcements (BAAs) on SAM.gov and Grants.gov, or as calls under the Artificial Intelligence Exploration (AIE) program announcement (DARPA-PA-22-02).
Answer the Heilmeier Catechism: Proposals must articulate what they aim to do, current limitations, what is novel, who cares, risks and payoffs, costs, and quantifiable milestone exams.
Submit an Executive Summary: Many BAAs encourage an initial 3- to 5-page executive summary. Program Managers provide feedback on whether the concept warrants a full proposal.
Develop Technical and Cost Volumes: Full submissions require a detailed Statement of Work (SOW), milestone verification plans, key personnel qualifications, and an auditable cost volume.
Research Security Compliance: Performers must comply with DARPA’s Countering Foreign Influence Program (CFIP), disclosing foreign appointments, international funding, and institutional conflicts of interest.

Typical timeline phases (no calendar dates as facts)

DARPA AI solicitations operate across distinct execution phases depending on the award vehicle:

Standard BAA Procurement Pathway:
- Phase 0 (Solicitation): BAA publication and Proposers Day briefing, followed by a 45- to 60-day proposal drafting window.
- Phase 1 (Proof-of-Concept): An initial performance period (typically 12 to 18 months) focusing on core theoretical development, synthetic benchmarking, and algorithm design.
- Phase 2 (Prototype Integration): An option phase (typically 18 to 24 months) awarded upon milestone verification, focusing on real-world testbed integration and stress-testing.
Artificial Intelligence Exploration (AIE) Fast-Track Pathway:
- Rapid Contracting Window: AIE opportunities move from announcement to project kickoff within an accelerated 90-day procurement window.
- Feasibility Sprint: AIE contracts validate concept feasibility within an 18-month maximum period, structured as a Phase 1 feasibility study with a Phase 2 option.

Common mistakes

Proposing Incremental Engineering: DARPA does not fund derivative improvements to commercial software. Proposals must promise transformative, order-of-magnitude advances.
Neglecting Explicit Milestones: DARPA contracts are milestone-driven. Proposals lacking quantifiable, falsifiable metrics are rejected during technical review.
Failing to Align with the Heilmeier Catechism: Submitting open-ended scientific narratives rather than structured risk-payoff frameworks with clear transition paths.
Misjudging Export Control Requirements: Failing to establish institutional data security protocols for dual-use artificial intelligence code and defense-sensitive technical data.
Non-Compliant Cost Volumes: Submitting non-standard overhead rates or misclassifying institutional indirect costs under FAR or Other Transaction rules.

Related programs

Research administrators evaluating federal artificial intelligence opportunities should compare DARPA’s mechanisms with complementary military and federal funding channels:

Program / Mechanism	Managing Entity	Primary Scope & Objective	Award Structure
DARPA AI Next / AIE	DARPA (I2O, DSO, STO)	Third-wave contextual AI, feasibility sprints, and hardware efficiency	18-month AIE sprints; multi-year BAA FAR contracts and Other Transactions
DARPA I2O Office BAA	DARPA Information Innovation Office	Cyber resilience, mega-system modeling, and foundational AI innovation	Multi-phase contracts and grants managed under office-wide announcements
ONR Long-Range BAA	Office of Naval Research	Naval-relevant maritime AI, ocean robotics, and machine perception (see ONR BAA guide)	Annual open solicitations for university and industry basic/applied research
AFOSR AI Research	Air Force Office of Scientific Research	Basic science (6.1) in autonomous reasoning, cognition, and machine intelligence	Standard academic grants and university research initiatives
NSF National AI Institutes	National Science Foundation	Long-term collaborative academic research across civilian and scientific sectors	Five-year cooperative agreements with university consortiums

When comparing foundational paradigms across computational disciplines, researchers frequently reference machine learning vs artificial intelligence conceptual divisions to determine appropriate funding alignment.

FAQ

What is the difference between DARPA AI Next and the AIE program?

The AI Next campaign is DARPA’s overarching $2 billion strategic investment portfolio spanning more than 60 research programs. The Artificial Intelligence Exploration (AIE) program is a specific fast-track funding mechanism under AI Next that solicits, awards, and launches feasibility projects within 90 days, with 18-month execution windows.

Do universities need security clearances to apply for DARPA AI funding?

No. Most university-led research funded under DARPA AI BAAs is fundamental research (Budget Activity 6.1) and is completely unclassified. Performers do not require facility or personnel clearances unless a specific solicitation explicitly indicates classified performance tracks.

What is an Other Transaction (OT) agreement in DARPA AI awards?

An Other Transaction (OT) is a flexible statutory procurement authority exempt from the Federal Acquisition Regulation (FAR). OTs are frequently used in DARPA AI and AIE solicitations to streamline intellectual property terms and attract commercial and academic innovators who do not operate federal cost-accounting systems.

How does DARPA evaluate AI proposals?

DARPA evaluates proposals on scientific and technical merit, relevance to national security, researcher qualifications, cost realism, and adherence to the Heilmeier Catechism. Reviewers assess the feasibility of stated technical milestones rather than traditional academic journal metrics.

Sources

Defense Advanced Research Projects Agency (DARPA). (2026). “AI Next Campaign.” https://www.darpa.mil/research/programs/ai-next-campaign (read 2026-11-07).
Defense Advanced Research Projects Agency (DARPA). (2026). “Information Processing Techniques Office / Information Innovation Office (I2O).” https://www.darpa.mil/about-us/offices/i2o (read 2026-11-07).