Autonomous trading agents, robo-advisors, and risk models make decisions that move markets. Project AIR provides the signed forensic evidence that proves what happened, why, and who approved it.
Capabilities
Every recommendation, execution, and risk assessment by an AI trading agent is captured as a signed intent capsule. Regulators can verify the complete decision chain.
Signed forensic chains provide the books-and-records evidence required under SEC Rule 17a-4 and FINRA 3110. Tamper-evident by design, not by policy.
When an AI advisory agent recommends a portfolio allocation, the signed chain proves what data it considered, what alternatives it evaluated, and why it chose the recommendation it did.
OWASP-mapped detectors flag unauthorized data access, privilege escalation, and exfiltration attempts. ASI03 Zero-Trust enforcement ensures agents only access authorized resources.
Containment Layer 3 requires JWT-verified human approval for high-value trades, large transfers, or account modifications. Bad tokens leave actions permanently halted.
When a research agent hands off to a trading agent, Layer 4 handoff protocol maintains cryptographic chain of custody with Sigstore Rekor counter-attestation.