At the Global Fintech Fest in Mumbai this week, two financial authorities described almost identical requirements for letting AI agents move money — without appearing to have compared notes. MAS managing director Chia Der Jiun described SAFR (Safeguards for Agentic Finance at Runtime), a white paper Singapore's financial industry published in July. Its core mechanism is a governance checkpoint sitting between an agent's decision and its execution, so no agentic action runs without first being declared, authorised, and assessed. In Chia's words, the safeguards mean "establishing an agent's identity and authority, evaluating agent actions against controls before execution, and maintaining a clear audit record." A day earlier, at the same event, India's National Payments Corporation of India (NPCI) said it was examining protocols to identify and authorise digital agents operating on UPI — the same three components: identity, pre-execution evaluation, and audit trail, arrived at independently by a different institution governing a different payment rail. Neither framework carries force of law. SAFR is explicitly voluntary — the paper states it does not constitute regulatory guidance and doesn't prescribe or anticipate future rulemaking. NPCI's work is still at the examination stage, not a published standard. So this isn't a case of one jurisdiction regulating where another hasn't; neither has actually made a rule yet. What's notable isn't the substance — pre-execution authorization checks and audit logging are the same pattern showing up across every serious agent-governance effort this year. What's notable is the redundancy: two of the world's more sophisticated payment infrastructures, working separately, landed on the same shape of solution before either had a binding mandate to do so. That convergence is itself evidence about where the actual pressure point is. Nobody drafting these frameworks is worried about whether an agent can technically execute a transaction. They're worried about the three-second gap between an agent deciding to act and that action being irreversible — and about who can say, after the fact, that the decision was authorized and correctly checked. The gap between the two announcements is instructive too: voluntary industry frameworks are outrunning statutory ones. SAFR exists as a published document; NPCI's UPI equivalent doesn't yet. If agentic payments scale before either becomes binding, the audit trail requirement will likely get built into infrastructure as a de facto standard well before any regulator forces the issue — which is either reassuring or exactly the kind of governance-by-default that makes oversight harder to contest later.