Liability and regulation
Coverage marks: verified means corroborated by multiple sources or a primary document; thin means a single secondary source, or an emerging item whose terms are not yet public. Where no rule exists, the record says so; absence of a rule is itself the finding. General information, not legal advice.
Sourcing caveat: Visa’s rulebook text itself is not public; Visa records rest on convergent legal analyses quoting it, not on the primary text. The UK consultation-response record is a respondent’s analysis, explicitly labeled as such, not an official position.
Global
2025-05-06 — Coinbase (x402) [thin]
Rule: The x402 protocol's exact scheme is a push payment that is irreversible once executed; a settled exact transfer is reversed only if the recipient sends funds back. The newer auth-capture scheme adds refund semantics. x402 commonly settles in USDC.
Impact: On x402, the settlement scheme chosen decides whether any remedy exists: the exact scheme leaves the payer with only the recipient's goodwill, while auth-capture adds refund semantics. Agent spending policies must therefore treat scheme selection as a liability decision, not a technical detail.
Sources: x402 exact scheme irreversible; auth-capture adds refund semantics
2025-09-16 — Google (AP2) [verified]
Rule: Google's Agent Payments Protocol (AP2) defines mandates as tamper-proof, cryptographically signed digital contracts backed by verifiable credentials. An Intent Mandate captures the user's initial instruction and provides auditable context for the interaction; a Cart Mandate locks in the exact items and price the user approved; a Payment Mandate authorizes payment against a specific instrument. Each mandate is issued by a wallet the user controls, signed with the user's key, and presented by the agent to the merchant or payment processor, which verifies the signature and checks mandate scope before charging. If a customer disputes a charge claiming they never authorized it, the merchant can present the mandate as evidence.
Impact: A signed mandate changes the evidence economics of an authorization dispute: instead of arguing from checkout logs about what a human clicked, the merchant presents a tamper-proof credential recording exactly what the user authorized. What a mandate does not prove is fulfillment: AP2 records authorization, not whether the deliverable matched the agreement.
Sources: AP2 mandates as verifiable proof of authorization, AP2 Intent, Cart and Payment Mandates explained, How AP2 verifiable credentials bind user intent
2025-10 — Visa (Trusted Agent Protocol) [thin]
Rule: Visa's Trusted Agent Protocol, published in October 2025 as an open standard, adds signed HTTP headers to the merchant request attesting the agent's identity, the user's consent, and the transaction intent. The attestation header carries the agent operator's identifier, a user-consent reference, and a payload hash binding the attestation to the specific cart. Merchants verify the signature against Visa's registry before routing the authorization.
Impact: Attestation headers let a merchant verify, before routing authorization, that the transaction comes from a registered agent acting under user consent for a specific cart. Because the protocol lives at the HTTP layer, it works across card brands, but it attests authorization, not the wisdom of the purchase.
Sources: Visa Trusted Agent Protocol attestation headers, Visa Trusted Agent Protocol introduced October 2025
2026-03 — Mastercard (Verifiable Intent) [thin]
Rule: Mastercard published Verifiable Intent with Google in March 2026, an SD-JWT credential chain whose constraint types include amount bounds and merchant allow-lists. Mastercard's Agentic Disputes mechanism uses the Verifiable Intent audit trail as evidence. As of April 2026 the system had no reported production deployments.
Impact: Constraint vocabularies (amount bounds, merchant allow-lists) let a merchant and an issuer check a transaction against stated limits at authorization time. Their liability value is unproven in production: as of April 2026 the system had no reported production deployments, so no dispute has yet tested a Verifiable Intent record.
Sources: Mastercard Verifiable Intent, SD-JWT constraints, zero production deployments as of April 2026, Mastercard Agentic Disputes uses Verifiable Intent audit trail as evidence
2026-04-14 — American Express [verified]
Rule: American Express announced Agent Purchase Protection on 14 April 2026 alongside its Agentic Commerce Experiences (ACE) developer kit. It protects eligible cardholders from charges tied to the error of a registered AI agent, provided the cardholder authorized the agent and the agent sent authenticated purchase intent, subject to terms and conditions. The announcement does not state how merchants are treated. Because Amex runs a closed-loop network, it can set cardholder protection without agreement from thousands of issuing banks.
Impact: Amex is the only major network to publish an explicit agent-error protection, and it is deliberately narrow: it covers eligible cardholders, not merchants by name, and only inside its registered-agent system. Merchants outside the Amex trust boundary get no equivalent commitment, and the four-party networks have published no equivalent guarantee.
Sources: Amex Agent Purchase Protection, 14 April 2026: scope and merchant gap, Amex ACE developer kit and Agent Purchase Protection
2026-04-18 — Visa [verified]
Rule: Visa's Core Rules of 18 April 2026 define an Agentic Transaction (an electronic commerce transaction undertaken on behalf of a cardholder, completed without direct interaction between the cardholder and the merchant) and state that a cardholder is responsible for any actions taken by an Agentic Payment Provider as part of an Agentic Transaction as if the cardholder initiated the transaction (Core Rules §4.1.24.10). Before acting, the provider must obtain the cardholder's acknowledgement of that responsibility and verify the cardholder's identity before storing a credential and again before acting on an instruction.
Impact: For issuers, acquirers and merchants, the cardholder is the counterparty on an agent purchase: the agent sits on the customer's side of the transaction, not the merchant's. Agent platforms are not merchants under this rule.
Sources: Legal analysis quoting Visa Core Rules §4.1.24.10, Legal analysis of the Visa agentic rules, Law360: Visa's Agentic Payment Rules Expose Compliance Tensions (28 May 2026)
2026-04-18 — Visa [verified]
Rule: The same 18 April 2026 Visa rulebook contains no dispute condition specific to an agent-initiated purchase (dispute codes run 10.1 to 13.9 without one). Agent-initiated purchases are routed into the existing card-not-present dispute process, and the update added no new consumer protections for agent purchases.
Impact: A merchant defending an agent-error dispute has no agent-specific reason code to answer: the buyer files an ordinary fraud or 'not as described' claim, and the merchant defends it with ordinary checkout evidence. Issuers and acquirers process agent-error volume through existing queues.
Sources: Legal analysis of Visa dispute codes 10.1-13.9, Analysis: Visa routed agent purchases into existing CNP disputes, Inc.: Visa's April 2026 rulebook defines agentic transactions but adds no agent dispute category
2026-04-18 — Visa [verified]
Rule: Under Visa's compelling-evidence (CE 3.0) framework, a fraud dispute (10.4) can be defeated where the merchant shows prior transactions on the same card or credential, matching device or fingerprint evidence, and other elements. A validly signed agent token (AP2, Verifiable Intent, or Trusted Agent Protocol) functions the same way: it defeats a fraud dispute by proving authorization. None of the three verifies whether the purchase itself was a good decision.
Impact: A validly signed agent token is strong fraud-dispute evidence for the merchant, but it proves authorization only, not that the purchase was the right one. An agent that was authorized to shop and then bought the wrong item still produces a consumer dispute, not an automatic merchant win.
Sources: Analysis: signed agent tokens defeat fraud disputes, verify authorization only, CE 3.0 requirements under the Visa Core Rules, April 18, 2026 edition
2026-05-19 — Mastercard [thin]
Rule: Mastercard's Chargeback Guide of 19 May 2026 contains none of the terms agentic, Agent Pay, or Know Your Agent. Fraud claims run under reason code 4837 and consumer disputes under 4853, and Mastercard's developer guidance confirms the existing frameworks apply to agent transactions.
Impact: Until Mastercard publishes agent-specific chargeback language, merchants and acquirers must map agent disputes onto fraud (4837) and consumer-dispute (4853) codes, and issuers have no published agent reason code to cite.
Sources: Legal analysis of the Mastercard Chargeback Guide, 19 May 2026
2026-07-07 — Card networks (general) [verified]
Rule: No formal liability shift exists for agent purchases: disputes follow existing chargeback rules. Under network rules, issuers generally bear the upfront cost of reimbursing the cardholder for an unauthorized transaction, but may shift that loss to the merchant's acquirer (and ultimately the merchant) through the chargeback process unless the merchant produces sufficient evidence that the transaction was properly authorized and authenticated. A purchase the agent got wrong, without fraud, fits neither the fraud code nor a dedicated agent-error code, and no network has published one.
Impact: In practice the issuer refunds the cardholder first and the loss is argued afterwards. Merchants absorb agent-error losses unless their evidence shows the customer authorized the agent and the agent stayed within scope. No published rule re-allocates the loss for a non-fraudulent agent error.
Sources: No formal liability shift for agent purchases; issuer refunds first, Issuer bears upfront reimbursement cost, shifts via chargeback unless merchant proves authorization
2026-07-07 — Merchants / agent platforms [thin]
Rule: The merchant that sells the goods and takes the payment is the merchant of record on an agent-completed purchase. OpenAI's production guidance and the Universal Commerce Protocol both state that the agent platform and payment service providers do not become the merchant of record, so the merchant handles refunds and chargebacks.
Impact: The merchant of record keeps refund and chargeback exposure on agent orders. An agent platform or payment service provider cannot be assumed to share that exposure unless a contract says so; platform terms should be checked before accepting agent traffic.
Sources: Merchant of record on agent purchases: agent platform and PSPs do not assume it
2026-09 — Visa / Mastercard (monitoring programs) [thin]
Rule: Visa and Mastercard count a dispute against the merchant's VAMP or ECM ratio regardless of whether an AI shopping agent or a human completed the purchase. Visa's VAMP threshold is 1.5% (from April 2026, down from 2.2%), counting both fraud reports and disputes; Mastercard's ECM triggers at a 1.5% ratio combined with 100 or more chargebacks in a month.
Impact: Agent-initiated disputes carry the same monitoring-program risk as human ones: a merchant's agent-order dispute volume counts toward Visa VAMP (1.5% threshold) and Mastercard ECM (1.5% ratio plus 100+ chargebacks in a month), with per-dispute fees, fines, reserves and possible account termination at stake.
Sources: Agent purchases count toward VAMP and ECM thresholds, Visa VAMP threshold 1.5% from April 2026
2026-09-01 — EMVCo [verified]
Rule: On 1 September 2026 EMVCo opened for comment a draft EMV Agentic Payments - Framework for Specifications, developed by its Agentic Payments Task Force. Its central proposal is Intent Services: a shared, interoperable layer that lets payment participants register, reference, retrieve and manage consumer-authorized intent before, during and after a transaction, complementing cryptographic assurance from existing industry solutions. The framework also explores Know Your Agent checks and Agentic Transaction Indicators. Public feedback was open through 30 September 2026. The draft does not address disputes or liability.
Impact: If adopted, Intent Services would give every party in the chain the same reference point for what was authorized, which is the precondition for any future liability rule: disputes could cite a shared intent record instead of each party's own logs. The draft itself deliberately leaves disputes and liability unaddressed, so it creates no liability rule today.
Sources: EMVCo drafts framework for agentic card payments, FIDO Alliance: EMVCo drafts framework for agentic card payments, EMVCo framework introduces Intent Services, does not address disputes or liability
2026-09-30 — Mastercard [verified]
Rule: Mastercard published the Agent Pay Trust Framework on 30 September 2026, combining identity, intent, controls, trusted execution and intelligence. Its white paper frames the governing questions as who is acting, on whose behalf, with what authority, under what limits, and with what accountability if something goes wrong. It introduces Agentic Disputes, which uses the Verifiable Intent audit trail as evidence, plus Know Your Agent checks and an Agent Pay Ready mark for participants meeting program requirements. Mastercard has not disclosed how losses on disputed agent purchases are split.
Impact: The Trust Framework gives the chain shared vocabulary (identity, intent, controls, execution, intelligence) and an evidence trail, but it does not publish how losses on disputed agent purchases are split between issuers, merchants and agent platforms. That split remains undisclosed and therefore unpriced.
Sources: Mastercard agentic commerce tracker: Trust Framework, Agentic Disputes, undisclosed loss split, Mastercard Agent Pay controls and Trust Framework, 30 September 2026
2026-09-30 — Mastercard [verified]
Rule: On 30 September 2026 Mastercard announced a probability score intended to estimate whether a transaction was initiated by an AI agent, tested in the United States. Mastercard has not disclosed the score's accuracy, testing results, participating issuers, or a date for general availability.
Impact: A score that flags agent-initiated traffic helps issuers and merchants route and monitor agent orders, but it is a fraud-prevention signal, not a liability rule: it does not say who pays when an agent buys the wrong item.
Sources: Mastercard probability score announcement and undisclosed accuracy, Mastercard agentic commerce tracker: September 30 score
2026-10-13 — American Express [thin]
Rule: On 13 October 2026 American Express published its Intent Intelligence Framework. It issues a Proof of Intent Token that cryptographically encodes the consumer's purchase constraints (merchant, budget, product specifications) approved before the agent acts. A transaction that does not match those constraints can be declined before it clears; if a dispute occurs, the Intent ID and Proof of Intent Token provide the evidentiary basis for resolution.
Impact: Pre-clearing decline moves liability prevention before settlement: a transaction that does not match captured intent can be stopped rather than disputed afterwards. The evidentiary value depends entirely on the Intent ID and Proof of Intent Token record, which only exists for transactions inside the Amex system.
Sources: Amex Intent Intelligence Framework and Proof of Intent Token, Amex Intent Intelligence Framework: registered agent plus authenticated intent
United States
2025-04 — United States (UETA) [verified]
Rule: The Uniform Electronic Transactions Act (UETA), adopted in 49 states and DC (New York excepted), defines an electronic agent as a computer program used independently to initiate an action without review by an individual (Section 2(6)). Its prefatory note states that electronic agents programmed and used by people will bind the user of the machine, and Section 14 provides that contracts form through agent interaction even if no individual was aware of or reviewed the agents' actions. The commentary treats the AI as a tool: the employer of a tool is responsible for the results obtained by its use, since the tool has no independent volition. Section 10's error framework covers system changes and user mistakes, not model failures such as hallucinations or prompt misinterpretation, and no reported case law exists for AI agent liability in erroneous transactions.
Impact: Contract formation law already attributes an agent's transactions to the person who deployed it, so a business cannot argue that no human reviewed the purchase. What contract law does not answer is whether the terms are fair, whether an agent that exceeded its instructions bound its principal, or how a model failure (hallucination, misread instruction) fits an error framework written for system changes and user mistakes.
Sources: UETA electronic agent doctrine and the model-failure gap, UETA and E-SIGN: agent contracts enforceable, fairness and ultra vires acts unresolved
2025-07 — United States (GENIUS Act) [verified]
Rule: The GENIUS Act, signed into law in July 2025, created a federal framework for payment stablecoins, requiring 1:1 reserves in liquid assets and regular third-party audits, with implementing rules still being finalized. It created no chargeback mechanism. Stablecoin payments are push transactions: once confirmed on-chain they are final, with no chargeback window, and fraud vectors shift to address poisoning, approval fraud on malicious contracts, and off-chain dispute friction.
Impact: Stablecoin settlement removes the chargeback safety net entirely: an agent that pays the wrong payee or the wrong amount on a stablecoin rail has no network dispute process to reverse it. Issuer regulation (reserves, attestations) governs the coin, not the transaction's correctness, so recourse must be built contractually before settlement, not after.
Sources: GENIUS Act framework, no chargeback mechanism created, GENIUS Act reserve and audit standards for payment stablecoins, Stablecoin push finality and relocated fraud vectors
2026-06-04 — Vercel (contract terms) [verified]
Rule: Vercel's terms update of 4 June 2026 provides that if a customer enables a third-party AI tool and grants it access to their account (API key, credentials, or repository access), the customer authorizes and agrees to be legally bound by the actions taken on their behalf by those third-party tools, and is responsible for any costs incurred through those tools' use of the services. The terms designate any connected third-party tool as an electronic agent as defined in UETA.
Impact: Platform contracts are already writing the UETA doctrine in before courts impose it: a business that connects a third-party agent accepts its actions and its costs by contract. Expect agent-platform and vendor terms, not statutes, to be the first place liability is actually allocated.
Sources: Vercel terms classify third-party AI tools as electronic agents; billing falls on the developer
2026-09 — United States (Regulation E / Regulation Z) [verified]
Rule: Regulation E (12 CFR 1005.2(m)) defines an unauthorized electronic fund transfer to exclude a transfer initiated by a person who was furnished the access device by the consumer, unless the consumer has notified the financial institution that transfers by that person are no longer authorized. The CFPB's official commentary states that if a consumer furnishes an access device and grants authority to a person who exceeds the authority given, the consumer is fully liable unless access has been revoked with the institution. Most agent checkout flows today run on stored credit-card tokens, which fall under Regulation Z and the Fair Credit Billing Act instead, not Regulation E, and Reg Z's unauthorized-use definition is worded differently with a $50 consumer liability cap. How courts treat furnishing to software rather than to a person is unresolved.
Impact: The debit and credit rails answer the furnished-credential question differently, and that difference decides which protection applies to an agent payment. On debit-linked agent payments, the furnished-device exclusion can leave the consumer fully liable even where the agent exceeded its instructions; on credit-card agent flows, Regulation Z's unauthorized-use definition and $50 cap apply instead. Whether courts treat furnishing credentials to software the same as furnishing them to a person is unresolved.
Sources: Regulation E furnished-device rule and the Reg Z distinction for agent payments, Regulation E commentary: consumer fully liable where the furnished person exceeds authority, Regulation E furnished-device exclusion explained
2026-09-29 — United States (Federal Reserve) [verified]
Rule: Federal Reserve Governor Christopher J. Waller, speaking at Sibos 2026 in Miami on 29 September 2026 in a speech titled Payments in the Age of AI Agents, distinguished the agent-assisted model (consumer retains control of decisions and payments) from the agent-delegated model (consumer grants an agent authority to shop and pay within constraints). He identified authentication, liability and fraud as the key challenges, called trust the biggest barrier to scaling agentic commerce, particularly the agent-delegated model, and noted that payment authentication must establish not only whether a buyer is authorized to pay but whether an agent has authority to pay on the buyer's behalf. He also revealed the Federal Reserve hosted an industry roundtable on agentic commerce earlier in the summer of 2026.
Impact: The US central bank has put liability on the official policy agenda without proposing a rule: authentication, liability and fraud are named as the three barriers to scaling agent-delegated payments, and trust is named the biggest barrier. Industry rule-writing (networks, platforms) is therefore the operative arena for now, not legislation.
Sources: Waller Sibos speech: authentication, liability, fraud as the three barriers, Waller: trust the biggest barrier to scaling agent-delegated commerce, Waller speech highlights and Fed industry roundtable
2026-10 — United States (CFPB / Federal Reserve) [thin]
Rule: As of 2026, no federal rule addresses agent-initiated purchases specifically. The CFPB and the Federal Reserve have remained silent on an agent-specific liability rule, and the administration has signaled a preference for innovation over immediate regulation, leaving the industry without a clear federal rulebook for agent-error losses.
Impact: No federal rule currently answers who pays when an agent buys the wrong thing, so US liability is decided by the combination of network rules, contract terms, and existing consumer-protection statutes (Regulation E, Regulation Z, UETA) rather than by any agent-specific regulation.
Sources: No federal agent-specific liability rule; CFPB and Fed silent
European Union
2018 — European Union (PSD2) [verified]
Rule: Under PSD2, in force since 2018, Article 72(2) provides that the recorded use of a payment instrument is in itself not necessarily sufficient to prove that the payer authorized the transaction. Article 73(1) requires the payer's provider to refund an unauthorized transaction immediately, and in any event no later than the end of the following business day, with a narrow exception where the provider has reasonable grounds to suspect fraud and reports those grounds in writing to the national authority.
Impact: In the EU, the burden of proof sits with the payment service provider, not the payer: a PSP cannot defeat an unauthorized-payment claim merely by showing the instrument was used. For agent disputes this matters twice over: the PSP must prove authorization, and a mandate or intent record is exactly the evidence that could do it.
Sources: PSD2 Articles 72(2) and 73(1): burden of proof and refund timing, PSR/PSD3 liability rules: PSP must prove authorization or fraud or gross negligence
2025-11 — European Union (PSD3 / PSR) [verified]
Rule: PSD3 and the Payment Services Regulation reached political agreement in November 2025 but are not yet applicable. Publication in the Official Journal is expected in December 2026, with the PSR applying about 21 months after entry into force (around late 2028) and Verification of Payee at about 27 months. Neither PSD3/PSR nor PCI DSS contains provisions written specifically for AI agents, and how a customer's authorization is passed to an agent is still being worked out. PSD3/PSR apply to payment services in the EU, so authentication, fraud liability and monitoring rules still apply when a payment is made on a customer's behalf.
Impact: Europe's next payment rulebook will not settle agent disputes on arrival: it contains no agent-specific provisions, and its application is years away. Until then, agent payments in the EU run on PSD2 plus network rules, with the authorization-delegation question still open.
Sources: PSD3/PSR timeline and absence of agent-specific provisions, PSD3/PSR political agreement November 2025, not yet applicable
2026-07-01 — European Union (MiCA) [thin]
Rule: MiCA's transitional period ended on 1 July 2026. MiCA provides the EU rulebook for crypto-asset issuers and services, but none of the three major frameworks reaching enforcement in mid-2026 (MiCA transition end, GENIUS Act implementation, EU AI Act high-risk obligations) contains provisions for autonomous machine-to-machine transactions.
Impact: For EU stablecoin settlement, MiCA's transitional period ending 1 July 2026 is the date after which the EU stablecoin rulebook fully applies, but like the GENIUS Act it contains no provisions for autonomous machine-to-machine transactions: an agent paying in stablecoins in the EU faces the same finality-without-chargeback position as in the US.
Sources: MiCA transitional period ends 1 July 2026; no provisions for autonomous machine transactions
2026-10 — European Union (AI Act) [verified]
Rule: The EU AI Act, the most comprehensive AI regulation on the books, does not address autonomous purchasing or payment liability; it focuses on AI risk classification and transparency requirements. The EU's AI Omnibus, formally adopted in mid-2026, moved the high-risk system deadlines to 2 December 2027 for stand-alone systems and 2 August 2028 for systems embedded in regulated products.
Impact: The EU's flagship AI law offers no liability answer for agent purchases: it governs AI risk classification and transparency, not who pays when an agent buys wrong. Payment liability in the EU therefore stays with PSD2, network rules and contract, not with the AI Act.
Sources: EU AI Act does not address autonomous purchasing or payment liability
United Kingdom
2026-06-30 — United Kingdom (Bank of England) [verified]
Rule: Bank of England Deputy Governor for Financial Stability Sarah Breeden, speaking at the European Central Bank Forum on Central Banking in Sintra on 30 June 2026, said the biggest issues for regulation and industry standards are likely to be how users securely give consent and authorization to agents (especially for multiple transactions), how disputes are settled and liability assigned for erroneous or fraudulent transactions, and how authorities avoid fragmentation and walled gardens as firms develop competing agent protocols. She added that existing technology-agnostic frameworks were not built to contemplate autonomous agents, and relying on a human in the loop for all agent actions is unlikely to be realistic. The Bank is leading a public-private partnership to design the next generation of UK retail payments infrastructure to support agentic payment journeys.
Impact: The Bank of England has named the same three questions as the Fed (consent, disputes/liability, fragmentation) and is building the next-generation retail payments infrastructure with agentic journeys in mind, but it has proposed no liability rule. UK firms should expect infrastructure and standards work first, liability rules later.
Sources: Breeden Sintra speech (Bank of England, primary text), Breeden speech: consent, liability, fragmentation as the biggest issues
2026-07-14 — United Kingdom (HM Treasury) [verified]
Rule: On 14 July 2026 HM Treasury published its consultation on modernising payment services regulation, the most significant restructuring since the Payment Services Regulations 2017 and Electronic Money Regulations 2011. It proposes moving technical and firm-facing requirements into FCA rules while keeping core consumer protections and perimeter provisions in statute, a single set of regulated activities covering fiat and tokenised payment services (including UK-issued stablecoins), and asks for views on how authentication, consent and liability rules must adapt for agentic payments. The consultation closed on 6 October 2026; the current plan is a Statutory Instrument during 2027-28 with FCA consultations in parallel.
Impact: The UK has opened the question but answered nothing: the consultation asks whether consent, authentication and liability rules need updating for agentic payments, and legislation is not expected until 2027 or 2028. UK liability therefore runs on the existing Payment Services Regulations 2017 plus network rules until then.
Sources: HMT consultation: overhaul, tokenised and agentic payments, closes 6 October 2026, HMT consultation executive summary: FCA-led model, agentic AI framework, HMT consultation: SI planned 2027-28, FCA framework in parallel
2026-10 — United Kingdom (consultation response) [thin]
Rule: In a response to the HM Treasury consultation, analysis of the Payment Services Regulations 2017 notes that regulations 76 and 77 allocate the loss on an unauthorized payment but do not reach a payment inside the delegation that the consumer did not intend, because the agent erred, was misled by content it read, or was changed by its provider after delegation. Regulation 79 may refund a card payment initiated through the payee where the authorization named no exact amount and the amount exceeded what the payer could reasonably have expected, on request within 8 weeks; an account payment the agent initiates has no such route. Section 75 of the Consumer Credit Act 1974 reaches a credit-card issuer only, for supplier misrepresentation or breach of contract, never for an item priced at £100 or less. The response proposes FCA rules setting what the record of an agent's payment must show, and Know Your Agent covering the agent's state, not just its identity.
Impact: This is a respondent's analysis, not an official position, but it identifies the precise UK gap: the existing unauthorized-payment rules do not reach a payment that stayed inside the delegation but was not intended. If the FCA adopts the proposed evidence-record and Know Your Agent rules, mandate and intent records would become regulatory evidence, not just network practice.
Sources: Consultation response: PSR regulations 76/77 gap, regulation 79, section 75 limits (PDF), Consultation response summary: evidence record and Know Your Agent proposals