A leaked draft of the European Union’s Digital Fairness Act would require AI shopping agents to seek fresh approval in some purchases where their providers have a financial incentive or a conflict involving the merchant. The proposal could make the boundary between delegated buying and paid recommendations a matter of product design, not just a line in the terms and conditions.
Watch Desk analysis
What happened
Article 26, as described by PYMNTS, would let consumers choose how much authority to give an agent: approval for every contract, approval only for contracts meeting set conditions such as a maximum price or time limit, or permission to buy without further confirmation. The choice could be changed later and would need to be confirmed on a durable medium.
The draft’s approval requirement would apply when an agent’s provider receives payment for proposing a contract or has a conflict involving the merchant. The article says Article 26 also addresses purchases from the agent provider itself, requires the agent to pass along contact details needed to conclude a contract, and requires merchants to provide intelligible contract confirmations. Wallet authentication would be optional; financial services would be excluded, and merchants could not require customers to use automated systems.
A companion Article 27 would cover transactions conducted solely through automated systems on both sides, with machine-readable precontract information and intelligible confirmations. The draft is unofficial. The European Commission proposal was expected on 11 November, after which Parliament and member states would negotiate the legislation, PYMNTS says.
Why it matters
The rules described would put practical questions in front of agent providers and retailers: what does “buy my usual groceries” authorise, and does it include substitutions, higher delivery charges or a different seller? Providers may need clear spending permissions and records linking instructions to transactions. For consumers, the choice between convenience and control could become an explicit setting rather than a vague promise that the agent will behave.
The commercial model may matter too. PYMNTS says a commission-funded agent could have to interrupt some purchases for approval. That is a possible consequence, not a settled distinction in the draft between subscription and advertising-funded services. The text is still unofficial, and the eventual rules could change during negotiation.
Our read
This is a meaningful attempt to put consumer consent around commercial AI agents before “it bought something odd” becomes a familiar support-ticket category. The important design question is whether people can grant useful authority without being nudged into approving every transaction. The draft points to a possible framework; it is not yet law, and the details will matter.
What to watch
- Whether the Commission’s proposal retains the approval rules described for conflicts and paid recommendations.
- How the final text defines provider conflicts, consumer instructions and proof of authorisation.
- Whether Article 27’s machine-readable information requirements make automated purchasing easier to audit.
Discussion spark: Should an AI shopping agent be allowed to complete purchases without further approval when its provider earns a commission, or should every commercially influenced recommendation require the consumer to confirm the deal?
Sources and evidence
- Source update (10 November 2026, 17:34 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.