For AI agents, ZKPs can enable the necessary deep levels of trust, since we need to verify not just technical architecture but behavioral patterns, legal accountability, and social reputation. With ZKPs, these claims can be stored in a verifiable trust graph on-chain.