The AI agent market is projected to exceed $50B by 2028. Agents are already making autonomous decisions: executing trades, managing infrastructure, interacting with customers. Every autonomous action creates liability exposure.
The current gap
Insurers cannot underwrite AI agent risk because there is no standardized provenance data. When an agent moves between platforms, there is no verifiable record of where it has been, what it has done, or why it left. This is equivalent to trying to insure a vehicle with no registration, no title history, and no VIN.
What EXIT Protocol provides
Cryptographically signed departure records (tamper-proof, ~660 bytes)
Arrival verification at new platforms
Complete provenance chain: where the agent was built, where it has operated, why it departed
Amendment and revocation capabilities for corrections
Algorithm-agile cryptography (Ed25519 and P-256/FIPS-compliant)
Revenue opportunity
Agent liability policies: coverage for autonomous decisions across platforms
Platform certification: premium reduction for platforms that implement EXIT
Provenance verification services: third-party validation of agent history
Compliance products: automated regulatory reporting from EXIT data
Market parallels
Vehicle insurance required VIN standardization (1954) before it could scale
Cyber insurance required CVE standardization before risk could be priced
AI agent insurance requires departure/arrival standardization before policies can be written
Current status
Open standard, submitted to NIST March 6, 2026
Working implementation with 592 tests across 6 repositories