Six global banks — ING Group, NatWest Group, ASB Bank, Bank of America, Capital One and Commonwealth Bank — have published a set of joint principles for trusted agentic commerce. The document, reported by The Paypers, is the most coordinated attempt yet by incumbent banks to shape how AI agents initiate payments on their customers’ behalf, rather than watching crypto-native rails define the rules first.
The significance is not the content of any single principle but the list of signatories. Banks in the Netherlands, the UK, New Zealand, the US and Australia rarely publish anything jointly about a technology this early. That they have done so for agentic commerce signals that agent-initiated payments are no longer a fringe experiment from their perspective — it is a competitive threat with a standards vacuum they intend to fill.
What the Banks Are Actually Doing
The six institutions have laid out shared principles covering trusted agentic commerce — the framework under which an AI agent acting for a consumer or business can initiate, authorize and settle payments. Details remain high-level, as first iterations of bank consortium documents usually are, but the direction is clear: establish authorization boundaries, accountability and consumer protection before agents transact at scale inside bank rails.
The timing matters. In the past week alone, Mastercard and Danske Bank cleared Denmark’s first agent-initiated payment, and card-network agent rails are multiplying alongside stablecoin-based protocols. Banks that sit behind card and account-to-account infrastructure need a common position on liability when a machine, not a human, clicks pay. These principles are the opening bid in that negotiation.
The Stablecoin Angle
The banks’ move lands in a market where the most active agent-payment infrastructure is already running on stablecoins. x402 has shipped implementations across multiple chains, Coinbase Agent Payments and Circle’s gas-free facilitator model let agents spend USDC without managing gas tokens, and platforms like Skyfire and Payman are building commercial agent wallets on stablecoin balances.\n That creates a two-track contest. The bank track offers regulated accounts, existing consumer trust and — now — a shared governance framework. The stablecoin track offers 24/7 settlement, programmability and no requirement that six institutions agree on anything before shipping. The stablecoin track is currently faster at protocol level; the bank track is faster at liability level. Which advantage matters more depends on whether enterprise adoption of agents is gated by technology or by legal certainty.
Why Standards consortia Form Now
Standards bodies typically form when a technology is too successful to ignore but not yet locked in. Agentic payments fit that profile. Machine-initiated transactions raise questions that human payments answered decades ago: who bears loss when an agent misfires, how a customer revokes authority, how merchants verify that an agent is genuinely mandated by the wallet owner.
The crypto-native answer has been cryptographic: signed authorizations, spending caps in smart contracts, facilitator screening of counterparty addresses. The bank answer will likely be contractual and regulatory: mandates, chargeback analogues, supervisory oversight. Neither approach is complete, and the open question is whether the two tracks converge — for example, banks settling agent transactions in stablecoins under their own principles — or harden into separate ecosystems.
What to Watch
Watch for three things. First, whether card networks and payment schemes endorse or absorb these principles — Mastercard and Visa both have agent initiatives in flight, and their posture will determine whether the bank framework becomes an industry standard or a bank-only position. Second, whether the principles explicitly address stablecoin settlement or ignore it, which will signal whether incumbents see crypto rails as compatible infrastructure or as competition. Third, whether regulators in the signatories’ home jurisdictions reference the principles in guidance — that would convert a voluntary document into de facto compliance expectations for anyone building agent payments in those markets.