Entrusting a purchase to an artificial intelligence agent that searches, compares, negotiates and pays on the customer’s behalf: that is the promise of so-called “agentic” commerce. In September 2026, six banks published a document proposing five common principles to the entire sector to frame this practice: transparency, security, privacy and data protection, choice, and interoperability. The text was presented by Fintechnews Switzerland.
Who signed it, and what is the text’s status
The document is co-signed by ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING Group and NatWest Group. It is a discussion paper from private actors: according to the report, its authors believe the industry “should adopt” these principles so that the ecosystem works safely and efficiently for all participants.
It should therefore be read for what it is. It is neither a law, nor a technical standard, nor a decision by a supervisory authority. Nothing in the report indicates that it creates an obligation, including for its signatories, and the source mentions neither an implementation timetable nor an enforcement mechanism. Its authors intend it for several audiences: AI agent providers, technology platforms, payment networks and providers, but also industry associations and regulators, who could use it to “inform discussions” on governance, consumer protection and market development. A contribution to the debate, then, and not its conclusion.
The five principles in detail
- Transparency. All parties should know when an AI agent is taking part in a transaction and on whose behalf it is acting. Consumers and merchants should understand how the agent ranks options and makes its decisions, as well as how their data is collected and used. The document also flags risks of conflicts of interest.
- Security. The text anticipates more scams, fraud and disputes, as well as diverging expectations about the product purchased or about liability when an agent exceeds its mandate. It recommends secure and auditable methods for entering payment data and authorising purchases, and asks that consumers be able to view and manage the delegations granted to an agent.
- Privacy and data. Conversational queries, decision logs, intent mandates, purchase details: agentic commerce generates new data. Providers should keep auditable records of instructions, authentication and decisions, access only the data that is necessary, and make any additional use or sharing subject to consent.
- Choice. The document warns against restrictions on access to platforms and the obligation to use affiliated services. Consumers and merchants should be able to choose their services without unreasonable restriction, and the identity of the merchant responsible for the sale should be clear.
- Interoperability. Fragmentation across agents, payment methods and protocols risks reducing network effects and making integration more expensive. The authors propose building common connectivity for core functions, while leaving room for innovation in value-added services.
What the principles commit to, and what they leave open
On reading, the document describes objectives more than means. It speaks of “appropriate” security and data protection standards without naming them, and of consent without specifying its form. The question of liability when an agent oversteps its authority is identified as a risk, but the report mentions no proposed rule for allocating it.
The text also leaves some leeway for the providers themselves: they should be able to choose which agentic services they accept, based on security criteria, commercial viability and legal grounds. The principle of choice therefore works both ways, which leaves banks and payment networks considerable latitude.
What about Switzerland?
No Swiss bank is among the signatories, and the source says nothing about how these principles fit with Swiss data protection law or with the rules applicable to payments in Switzerland. No reaction from a Swiss authority is reported. It is therefore not possible, on the basis of the information available, to say whether these principles go beyond the obligations already in force, overlap with them or depart from them.
In substance, several of the document’s themes, such as consent, data minimisation and the traceability of payment authorisations, touch on areas where the law often already sets requirements. That is precisely where their real scope will be decided: a voluntary principle only has an effect if it adds to what the law requires, or if it helps to interpret it.
A market still emerging
Agentic commerce is still at an early stage, the source points out, even though some players are already developing protocols so that agents can handle product discovery, decision-making and the transaction. According to a McKinsey study cited in the article, it could orchestrate between 3,000 and 5,000 billion dollars of global business-to-consumer retail commerce by 2030.
In Europe, a survey conducted in 2025 among consumers in France, Germany and the United Kingdom indicates that 84% use AI tools daily, and 38% to research products or decide on a purchase. Use remains mainly located upstream of the transaction: 63% use them to compare brands, models, prices and reviews, 55% to find out about a category or a product, and 46% to discover new products and find inspiration. Payment delegated to an agent, which is what these principles target, is therefore not yet the dominant use.
Finally, it should be noted that the Fintechnews Switzerland article does not link directly to the six banks’ document. The quotations and the details of the principles are taken from this report, not from the primary text.
