DARPA AI Programs

DARPA AI Programs represent the cutting edge of artificial intelligence research for national security applications. Through initiatives like AI Exploration (AIE) and AI Forward, DARPA invests in trustworthy AI systems with awards up to $1M for AIE (18-month rapid prototyping) and $10M+ for full programs. DARPA pioneered AI research for 60+ years and currently runs 70+ programs leveraging AI/ML technology.

For ZHC builders, DARPA offers unparalleled funding for high-risk AI research focused on trustworthy systems, human-AI teaming, and foundational AI theory. The agency values breakthrough approaches over incremental improvements and uses streamlined AIE contracts to achieve funding within 3 months of announcement. Projects must demonstrate clear paths to national security applications.

Program Categories

AI Exploration (AIE)

• Phase 1: Feasibility study (3-9 months)
• Phase 2: Proof of concept (9-15 months)
• Total: Max 18 months, $1M combined
• Speed: Award within 3 months of opportunity

Full AI Programs

• Duration: 2-5 years
• Funding: $2M-$20M+ per performer
• Scope: Comprehensive research programs

AI Forward Research Thrusts

• Foundational theory - Understanding limits and capabilities
• AI engineering - Predictably building real-world systems
• Human-AI teaming - Fluent, intuitive collaboration
• Trustworthy AI - Reliable, ethical, explainable systems

Priority Areas

• Contextual reasoning beyond pattern recognition
• Explainability and assurance for AI decisions
• Hybrid symbolic-neural approaches
• Meta-cognition and self-awareness in AI
• Adversarial robustness and security

Eligibility Requirements

• All performers - Companies, universities, FFRDCs, non-profits
• Security clearances - May be required for classified work
• US presence - Work performed primarily in US
• Unclassified proposals - Initial submissions must be unclassified

ZHC-Specific Fit

✅ Excellent for:

• Novel AI architectures and algorithms
• Explainable AI systems
• Human-AI interaction frameworks
• Adversarial ML and security
• Multi-agent coordination systems
• AI for autonomous decision-making

❌ Poor fit if:

• Incremental improvements to existing methods
• No clear national security relevance
• Purely commercial applications
• Standard machine learning implementations

AIE Application Process

Current Programs (2026)

• DARPA-PA-25-03 - AI Exploration opportunities (rolling)
• AI Forward - Trustworthy AI initiative
• EMHAT - Exploratory Models of Human-AI Teams
• FACT - Friction for Accountability in AI
• FoundSci - Foundation Models for Scientific Discovery

Proposal Components

• Technical approach and innovation
• Relevance to national security
• Team qualifications and facilities
• Milestone-based schedule
• Budget and cost breakdown
• Transition and impact strategy

• DARPA AI Programs
• DARPA Opportunities
• AIE Portal
• Email: darpa-pa-25-03@darpa.mil (AIE inquiries)
• SAM.gov - Federal registration
• Information Innovation Office - Primary AI office