Regulators: AI Agents Should Not Authorize Payments on Their Own

At GFF 2026 on September 25, 2026, three regulators answered the same question on the same day — whether AI agents should authorize payments — and gave the same answer: not on their own. Under the architecture described at the event, an agent's job is limited to determining user intent, while a separate authorization layer decides whether money actually moves.

That position had already surfaced a day earlier, when the AI chief of a Swiss bank said the same thing, according to the report. Neither the three regulators nor the Swiss bank are named in the material.

The account also cites Ajay Kumar Choudhary, non-executive chairman of NPCI, who spoke to the same point. He is quoted as saying that agents "may determine user intent," but the quotation is cut off in the source material, so the rest of his statement is not available.

As of the October 4, 2026 report, the matter rests at the level of public statements from regulators and a banking executive: agents may determine intent, but they do not hold the authority to authorize payments independently, and the decision on whether funds move stays with a separate authorization layer. The material does not indicate that any of the parties has published rule text, a formal framework, or a timeline. It also leaves out the reasoning behind the position, the types of payments it would cover, and how it would be enforced.