Primus CEO says AI needs verification, not blind trust

Xavier, CEO of Primus (PRIM), a decentralized data verification and computing network, said during CoinNess Night Live, our ongoing ask-me-anything program on Telegram, that even when tasks are completed by AI agents, users should put the results through as much verification as possible rather than checking only the final output and trusting it unconditionally. Xavier said Primus supports verifying AI-agent workflows from the process level, which helps improve transparency. Most AI agents operate through API calls, and Primus’ zkTLS technology can prove those calls were actually made, Xavier said. He added this kind of verification can help reduce AI misjudgments, manipulation and unexpected harm, and said building verifiable AI now to manage risk is a more practical approach than trying to predict scenarios 10 years from now.