
THE PROBLEM Autonomous AI agents now initiate real-world asset transactions — tokenised property, invoice finance, commodities — without human review at the signing layer. Existing guardrails sit above the agent, not between the agent and the wallet. A single manipulated prompt, a drifted policy, or a misconfigured permission scope can produce an irreversible on-chain transaction before any audit system fires. THE SOLUTION ProofBridge Liner is a pre-signing enforcement gateway purpose-built for tokenised RWA environments. It sits as a cryptographic policy layer between the agent's intent declaration and the wallet's signing operation. No transaction clears until it passes a structured policy evaluation cycle. Core mechanisms: 1. Intent Interception — Every agent-initiated payment or transfer request is captured before signing and decomposed into a structured metadata object: declared intent, target address, asset class, value threshold, counterparty risk tier. 2. Policy Gate Evaluation — Each request is evaluated against a YAML-defined policy ruleset (ALLOW / DENY / HUMAN_REVIEW / QUARANTINE). Rules are composable, auditable, and updateable without redeployment. 3. Cryptographic Audit Trail — Every evaluation event — pass or block — is logged with a tamper-evident hash chain. The audit log is structured for regulatory readability: a compliance officer or regulator can reconstruct the full decision chain for any transaction. 4. Circuit-Breaker Logic — Anomaly thresholds (velocity, value, counterparty deviation) trigger automatic suspension of the agent's signing authority pending human review. Recovery requires an explicit re-authorisation event.
19 May 2026