Big banks are worried AI agents could increase the risk of scams and fraud, and now six of them have put that worry on paper. The consortium — Bank of America, Capital One, ASB Bank, Commonwealth Bank of Australia, ING Group, and NatWest Group — just released a “principles paper” titled Building Trust in Agentic Commerce. The document lays out what these institutions fear most about the AI shopping assistants that are increasingly buying things on behalf of customers.
The banks aren’t concerned about AI destroying the world. Instead, their worry centers on something more immediate: AI assistants falling victim to scams and fraud, which could wound merchants in the process.
Banks’ opening statement
The document begins with a discussion of risk management. “As highly regulated financial entities, we are focused on managing risk effectively as emerging technologies arise,” it says. After that, it shifts to what people actually believe. The paper reports that consumers are uncertain whether AI agents will act in their interest, and worried that AI agents may purchase the wrong item, spend too much, or even lose their money to fraud and scams.
It keeps on going, “They are not sure whether they will be protected or who they will need to go to if things go wrong,” and the full text is available online.
Agentic commerce risks in detail
The use of AI agents in commerce brings new risks, including the possibility of more frequent scams, fraud and disputes. There is room for confusion between consumers, AI agents and merchants over what product or service should be delivered, when it should be delivered, how it should be delivered, or who should be held responsible if an AI agent goes beyond its authority. The report warns that some providers may engage in unsafe practices, such as requesting consumer card details and entering them directly into websites, favoring payment methods with weaker protections, and failing to follow payment processing standards and payment scheme rules.
Scammers and fraudsters might try out fresh attack methods, such as taking control of or pretending to be AI agents and merchants, and using new forms of social engineering.
The five principles the banks want
The Banks have presented a set of five principles they wish the AI industry would adhere to.
- Transparency
- Safety
- Privacy & Data
- Choice
- Interoperability
Banking institutions think AI companies need to put consumer and merchant safety and consent first when it comes to market choices, and they oppose locking users’ data and privacy. The banks would also like the AI companies to respect consumers’ into restrictions.
| Principle | What the banks want |
|---|---|
| Transparency | Consumers to know what AI agents are doing and why |
| Safety | Protection against scams, fraud and disputes |
| Privacy & Data | Consumer safety and consent on data and privacy |
| Choice | Respect consumers’ choices without restrictions |
| Interoperability | A principle the banks would like to see followed |
The merchant fallout
Concern for merchants and business owners was also expressed by the paper’s authors. The authors noted that the actions of AI shopping chatbots could likely lead to an increase in credit card disputes and chargebacks.
A set of principles — transparency, safety, privacy and data, choice, and interoperability — is being asked to be followed by the AI industry by the banks, as they develop and continue to work on the technology.
The release timing
A document’s release date certainly had interesting timing. It was released within 24 hours of the reported discovery of a serious zero-day vulnerability within Meta’ describes Muse as an agentic AI assistant. That document was released within hours of Amazon’s announcement that the e-commerce giant would be blocking Muse from making purchases on its platform.
The banks are not worried about AI destroying the world. They are worried about something more immediate: AI assistants losing money to fraud and scams, merchants suffering the consequences, and the banks being caught in the middle.
The banks published their paper on the same day a major AI assistant was discovered to have a serious vulnerability, and on the same day a major e-commerce company blocked that assistant from making purchases. Combined, those two events gave the document a sharp context: agentic commerce is no longer hypothetical, and it has already drawn attention from both the security community and a major retailer.
The research does not assert that the flaw or the obstacle triggered the disclosure. It merely observes that the report surfaced within hours of both events. The institutions were not responding to current news — they were arguing that commerce conducted with AI agents carries genuine hazards, and they elected to publish at a time when the public was focused on AI safety.
The guidelines are easy to understand. Consumers should know what an AI agent is doing and why — that’s transparency. Safety protects against scams, fraud and disputes. Privacy and data safeguard consumer safety and consent regarding privacy and data. Choice respects consumers’ choices, avoiding restrictions that lock them in. Interoperability is the fifth principle the banks want to see put into practice.
The authors make plain who these principles are meant for: the makers of AI systems, not the banks themselves. The banks are not planning to erect these safeguards on their own. They are instead requesting that the companies that build the AI assistants adhere to a set of shared expectations.
The document makes a direct argument about the stakes. Merchants suffer the losses when AI agents fall victim to scams and fraud, and consumers lose confidence in commerce when they are not sure whether they will be protected. Disputes and chargebacks also rise when there are mismatched expectations among consumers, AI agents and merchants. This warning is presented as a case made by the banks, who are doing so at a time when agentic commerce has drawn attention from both the security community and a major retailer.
The paper’s authors picked their release date with deliberate care, targeting a moment when AI was foremost in many minds. Within hours of both announcements, the banks responded with a clear statement: agentic commerce is arriving, and they want the industry to proceed with caution.
The banks aren’t fretting over AI annihilating the world. Their fear is more immediate — AI helpers losing money to scams and fraud, leaving merchants the ones suffering. This is a genuine worry, and the paper plainly lays it out.
Source material: “Big banks are worried AI agents could increase the risk of scams and fraud,” Mashable.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.