On 22 September NatWest, ING, Bank of America, Capital One, CBA and ASB published five principles for trusted agentic commerce. The clause I'd have led with is 2.3: providers should authorise purchase intents and payments through secure, auditable methods, "with any party holding a potential liability being able to require authentication". In the EU that party is the issuer, which refunds unauthorised transactions by default under Article 73 of PSD2.
The same paper admits the catch. Issuers may not have real-time access to the agent's identity, the merchant of record, the consumer's intent or the purchase details, and when I checked the published EMV 3DS specification I found no field that tells an ACS an agent is initiating the transaction. That leaves you holding the right to require authentication without the information to decide when to use it.
An agent indicator would help, but I'm no longer sure it's the signal that matters most. It tells you what is acting, not what the customer authorised it to do: this merchant, up to this amount, choosing the product itself, on authority that hasn't lapsed. EMVCo's draft framework, open for comment until 30 September, already treats the two separately. It proposes a way to register consumer-authorised intent now and leaves agentic transaction indicators for later work.
If you could put one of them in front of your ACS first, which would you choose: the agent indicator, or the intent record?
First posted on LinkedIn on 29 September 2026, where the discussion is. A week later I took the intent-record question further in Authorised to do what?
This post is from Out of band, the shorter arguments I make on LinkedIn between essays, kept here with their sources. For the long form, read the essays.