Six global banks have published a joint set of principles for agentic commerce, and their central message is that the technology is moving faster than the trust infrastructure around it. Bank of America, Capital One, ING, NatWest, ASB Bank and Commonwealth Bank of Australia released the framework on Sept. 22, aimed at guiding how AI agents should be allowed to transact.
Highlights#
The banks argue that consumer confidence, not capability, is the binding constraint. Shoppers are unsure whether agents will act in their interest, worry they will buy the wrong thing or overspend, and do not know who to turn to when a transaction goes wrong.
Fraud and dispute risks rise when autonomous software touches money, the banks say. They warn that some providers could steer toward payment methods with weaker safeguards, and that criminals could compromise or impersonate agents and merchants for new forms of social engineering.
When things do fail, liability is allocated unclearly across the value chain and dispute processes often exclude relevant parties, according to the report.
Bank of America's payments head Mark Monaco framed the challenge as building confidence through identity, authorization, fraud prevention, liability management and customer protection.
Why it matters#
The banks are describing the exact gap escrow and on-chain reputation close: enforceable spending limits, verifiable agent identity and a neutral record of what each agent actually did. As these principles harden into bank requirements, settlement rails that natively carry that evidence will have an edge.