September 22, 2026, (Inside AI) — A coalition of global banks, including NatWest, Bank of America, ING, ASB Bank, and Capital One, issued a stark warning on Tuesday: the rapid rise of AI shopping agents poses significant risks of scams, fraud, and data-privacy breaches. The group published a report outlining principles for the safe development of agentic commerce, just as technology giants like OpenAI, Anthropic, Google, and Meta aggressively promote AI chatbots that can select products and make purchases on behalf of consumers.
The banks' intervention highlights a growing tension between Silicon Valley's vision of effortless, AI-driven shopping and the financial sector's duty to protect customers. While consumers show enthusiasm for the convenience, the report notes they are largely in the dark about who bears liability when an autonomous agent makes a costly mistake or falls victim to a sophisticated scam.
Banks Demand Transparency And Interoperability
The report, which will be discussed with policymakers, proposes a series of safeguards. These include requiring clear disclosure when an AI agent is involved in a transaction, greater transparency into how those agents make decisions, and robust protections for customer data. The banks also insist that consumers and merchants must retain the freedom to choose which AI-powered e-commerce services they use, and that different systems should be interoperable to prevent lock-in.
"Consumers are unclear if AI will act in their interests," the report states. "They are concerned that AI agents may buy the wrong thing or spend too much - or even worse, lose their money to scams and fraud. They are not sure whether they will be protected or who they will need to go to if things go wrong."
Read: Pinterest Launches Visual Search Ads and AI Restyle Feature
The specific risks identified are not hypothetical. The report warns that AI agents could request customers' card details and enter them directly into websites, bypassing familiar bank security protocols. Worse, they might steer users toward payment methods that offer weaker protections, such as direct bank transfers or cryptocurrency, which lack the chargeback mechanisms of credit cards.
This warning arrives as retailers race to adapt. British retailer John Lewis reported in September that searches originating from AI agents had surged to 2.5% of its total, up from just 0.3% a year earlier. That eight-fold increase in twelve months underscores how quickly consumer behavior is shifting, even before mainstream adoption of dedicated AI shopping assistants.
The banks' principles echo concerns raised by consumer advocates and regulators globally. The U.S. Federal Trade Commission has already signaled interest in how AI agents might violate existing consumer protection laws, while the European Union's AI Act includes provisions for transparency in automated decision-making. However, the banks argue that current standards are insufficient for the unique challenges of agentic commerce, where an AI acts as an intermediary with access to sensitive financial data.
One core issue is liability. If an AI agent buys a counterfeit product or transfers funds to a scammer, who is responsible? The bank, the AI provider, the merchant, or the consumer? The report does not offer a definitive answer but calls for clear rules before widespread adoption. Without such clarity, banks fear a erosion of trust that could slow the very innovation they acknowledge is inevitable.
The coalition's move is also a strategic play. Banks risk being disintermediated if AI agents handle payments directly, bypassing card networks and traditional banking rails. By setting principles early, they aim to shape the regulatory landscape in their favor, ensuring that any new system includes the fraud detection and dispute resolution they provide.
Technology companies have not yet responded publicly to the banks' report. OpenAI and Google have both launched shopping features for their chatbots in recent months, with Google's Gemini now able to complete purchases via Google Pay. Anthropic's Claude and Meta's AI assistants are also experimenting with transactional capabilities. None have detailed how they will handle fraud liability or data privacy in agentic transactions.
The banks plan to engage with regulators in the U.S., U.K., and EU in the coming months. Their proposals could influence upcoming legislation, including the U.K.'s Digital Markets, Competition and Consumers Bill and potential updates to the EU's Payment Services Directive.
For now, the report serves as a shot across the bow. It acknowledges the promise of agentic commerce but insists that without guardrails, the technology could expose millions to financial harm. As one bank executive involved in the report put it, speaking on condition of anonymity, "We are not against AI shopping agents. We are against AI shopping agents that operate without accountability."
The coming year will test whether the tech industry and financial sector can collaborate on standards, or whether regulators will impose them. With AI agents already handling billions in transactions, the stakes could hardly be higher.
