ANSEN co-founder Zheng Li warns that a correctly authenticated AI agent can still make unauthorized decisions. Khaleej Times published his direct interview on October 3.
He calls for controls over individual actions. Permission to summarize alerts should not automatically authorize interventions in critical services.
Li distinguishes fast routine work from consequential decisions. Security policies and real-world interventions should remain subject to traceable human control.
His warning describes possible misconduct using valid credentials. The interview provides no independent effectiveness evidence for the safeguards he advocates.