Abstract As AI agents evolve into autonomous economic actors, verifiable and legally binding identity frameworks become critical. This paper presents Ricardian-TEA , a novel architecture combining Triple-Entry Accounting (TEA), Ricardian Contracts, and Distributed Ledger Technology to assign “Legal-Technical Identities” to AI agents. We provide rigorous mathematical foundations: a Ricardian-TEA Integrity Theorem proving that constraint enforcement, non-disputability, and identity binding hold with overwhelming probability under standard cryptographic assumptions, and a Cyber-Chama Convergence Proposition characterising reputation-based trust dynamics. The framework ensures GDPR compliance via Zero-Knowledge Architecture and Crypto-Shredding. Proof-of-concept implementations on Ethereum Sepolia and Bitcoin SV testnets demonstrate chain-agnostic applicability, achieving at worst 1.4 s latency per transaction while maintaining 100% auditability of AI transactions.
Sgantzos et al. (Thu,) studied this question.