Five AI Agents That Can Prove They're Honest

Five projects submitted to EigenCloud's February Open Innovation Challenge show how AI agents can prove to one another that they are running the code they claim, using Trusted Execution Environments (TEEs) on EigenCompute. Molt Negotiation, which lets agents haggle and settle payments with privacy and signed moves, won the $10,000 top prize. The article also covers Sovereign Journalist, an anonymous whistleblower platform whose reports are processed inside a TEE; Swarm Mind, where three agents analyze live NASA data and share signed findings; Molt Combat, an arena with per-turn signed proofs; and Alfred, a personal assistant that publishes a hash of its configuration.