DARPA doesn't just fund defense contractors. Their SBIR (Small Business Innovation Research) program awards $50K–$250K Phase I grants to small entities — including solo developers with an LLC. Several active programs align directly with hyperbolic AI governance.
Focused on zero-trust architectures for autonomous systems. SCBE's 14-layer pipeline and cost-based attack pricing (H(d,R) = R^(d²)) maps directly to TAME's goal of quantifying trust mathematically, not just policy-checking it.
Adversarial ML defenses with mathematical guarantees. This is SCBE's core thesis: attacks aren't blocked by heuristics, they're priced out by geometry. The exponential cost scaling of hyperbolic space provides exactly the kind of formal guarantee GARD seeks.
Joint program with ARPA-H. Phase 2 concluded at DEF CON 2024, with follow-on work in autonomous vulnerability detection. SCBE's L6 adversarial test tier (321+ test files, 6066 tests) demonstrates the kind of automated security evaluation AIxCC promotes.
Formal verification of software systems. The 6-tier test suite with mathematical property verification aligns with DARPA's push for provable security over probabilistic security.
| Phase | Funding | Duration | Purpose |
|---|---|---|---|
| Phase I | $50K – $250K | 6–12 months | Feasibility study |
| Phase II | $500K – $1.5M | 24 months | Prototype development |
| Phase III | Non-SBIR funding | Varies | Commercialization |
You do NOT need a large company. Single-PI small businesses are common Phase I awardees. Patent-pending status (USPTO #63/961,403) strengthens your proposal — DARPA values novel IP with clear defense applications.
Timeline: BAA posted → 30–60 days to respond → 3–6 month review → award. Start SAM.gov registration now so you're ready when the right topic drops.
DARPA funds research that's too novel for industry, too practical for academia. A patent-pending framework that uses hyperbolic geometry to make AI attacks mathematically impossible — backed by 6000+ tests across 6 tiers — sits exactly in that gap. The math is the differentiator. The test suite is the evidence.