Production-Deployed Trust Infrastructure for Autonomous AI Agents

A recent publication introduces MolTrust, a trust infrastructure for autonomous AI agents that has been deployed in production. It is based on W3C Verifiable Credentials 2.0 and Decentralized Identifiers v1.0, featuring on-chain anchoring on Base Layer 2. This system tackles the absence of a unified trust layer among autonomous agents, which currently number 69,000 and have completed 165 million transactions, amounting to a cumulative volume of 50 million USDC in a single marketplace. Regulatory bodies, including Singapore IMDA, NIST CAISI, and the EU AI Act, as well as prominent AI organizations like Anthropic and Google, recognize the necessity for a cryptographically verifiable trust framework. The architecture of MolTrust comprises four key components: identity, authorization, behavioral record, and portability, alongside a five-party accountability chain and the Agent Authorization Envelope (AAE). The paper can be found on arXiv with the identifier 2605.06738.