Banks exclude external agents
„External AI agents log in with the customer's credentials, and banks say they have tried to act on accounts even without customer instructions. Banks therefore need to check, on their own terms, which agents may access accounts.”
Counter-narrative
The ban also serves to protect deposits and customer relationships. US banks reportedly fear the exodus of cheap deposits. Apollo's chief economist states that assistants optimizing client funds could steer deposits toward higher-yield fintech accounts. On the fintech side, Self Labs responded to the bans with its Agent Vault product.
Carriers
JPMorgan / Jamie Dimon, Bank of America, Wells Fargo, Morgan Stanley, kanadai nagybankok (RBC, BMO, TD), OCC and Hong Kong banking supervision, Axios
Analysis
The narrative frames the issue as one of security, namely who has access to a customer account, but it is really about who owns the customer relationship. It is fed by the wave of agent bans, Bank of America's liability-shifting customer notice, and Know Your Agent solutions. Banks have a dual interest: it reduces their fraud liability and slows the outflow of cheap deposits. Its weak point is that, in the material reviewed, there is no public case data on agent actions taken without instruction, only bank-side detection. The likely outcome is not a ban but the authorization of identified, certified agents. The joint framework-building by Visa, Mastercard and Ant, and the Hong Kong requirement, point in this direction.
What's at stake: Who owns the customer relationship and the deposit if software decides on behalf of the customer?
Narrative sub-threads
Major banks ban external agents
Major US and Canadian banks are successively excluding third-party AI agents from customer accounts. Bank of America warns customers they may be liable if they share login credentials with an AI agent.
This branch refers to bans already implemented by banks themselves: specific institutions, specific agents (Muse, Instinct), and warnings sent to customers, not a standard or analysis.
↗ rising · evidence: moderateIndustry agent identification standards
Payment networks and identity providers are working on "Know Your Agent" frameworks and standards to verify agents and their authorization before transactions. American Express and LexisNexis have already launched their own solutions, but there is currently no consensus on aligning systems from Visa, Mastercard, and Ant.
Instead of bans, it relies on market, standards-based technical solutions for identifying and authorizing agents; carriers are payment networks and KYC companies.
↗ rising · evidence: moderateRegulators on agent authentication
Bank supervisors and central banks are beginning to mandate or examine whether agents are registered and authenticated. The OCC classifies simple payment agents as high risk; Hong Kong reportedly requires identity verification, and India is preparing an agent registry and a unified agent protocol.
The actors here are authorities, and the mechanism is supervisory expectations and registration, not banking business decisions or market standards.
↗ rising · evidence: moderateBlurring of consent and fraud liability
When software does the buying, the line between an approved payment and fraud blurs. Traditional authentication, such as a fingerprint, cannot be used, so it is an open question who bears the loss. In SEON's survey, 65 percent of respondents would blame the AI company if an agent committed fraud.
The evidence is a legal and conceptual analysis, plus a consumer survey on whose fault agent fraud is, rather than a ban or standard.
↗ rising · evidence: moderateFears for deposits and customer relationships
Alongside security arguments, business fear also emerges. Agents that optimize money may steer deposits toward higher-interest accounts, while platform agents insert themselves between the bank and the customer.
This branch prioritizes competitive dynamics and deposit sources over security arguments; economists, market reactions, and customer relations debates drive the discussion.
Sub-thread stories
Major banks ban external agents
JPMorgan blocks outside AI agents from customer accounts; Schwab also blocksretail customerDo not give an AI agent access to your account: if you do, every step the agent takes may be your responsibility.banking professionalIdentifying and blocking agent traffic, and handling customer consent, should be written into policy.fintech companyA provider offering an outside agent will have no direct entry to large-bank accounts and must negotiate an authorized access route.US big banks block third-party AI agents from accessing customer accountsretail customerDo not hand your banking login details to an AI agent, because banks may treat such access as prohibited or unauthorised activity.banking professionalRecognising agent traffic and tightening customer authentication rules is a direct operational task.fintech companyAgent-based services must find an authorised API channel instead of the screen- and password-based access that banks prohibit.Major U.S. banks block AI agents that log in with customers' credentialsretail customerBanking passwords and login details should not be given to AI assistants, because the bank prohibits it and the activity is hard to trace.banking professionalBanks already struggle to tell whether the account owner or an AI agent is carrying out an action, so identifying agent traffic is an urgent task.fintech companyAgent products built on sharing customers' credentials rest on a risky business model; official API access is needed.Canada's big banks, including RBC, BMO and TD, block third-party AI agentsretail customerDo not give out banking credentials to use a third-party AI agent for banking tasks, as the bank may block access.banking professionalDetecting and blocking agent access, and communicating with customers, are new tasks for digital channels.fintech companyAgent-based services must expect banks to block access and should seek official integration.JPMorgan, Amazon and Bank of America restrict external AI agents' accessretail customerAccording to the bank, giving an AI agent access to a bank account is at the customer's own risk, so login credentials should not be handed over.banking professionalAgent access needs a dedicated policy and customer communication; general terms of use are not enough.fintech companyAgent-based services may be blocked by large banks, so an official access solution is needed.
Industry agent identification standards
Ant International, Visa and Mastercard agree on common identity standard for AI agentsfintech companyAgentic payment products will have to be aligned with this identification standard, or they risk being left out of the networks.banking professionalBanks need to prepare to identify, authorise and handle disputes for transactions initiated by agents.retail customerThe prospect of an AI agent paying on a customer's behalf is approaching, so the limits of the mandate need to be set in advance.Visa, Mastercard and Ant seek to align AI agent verification, but disagree on the solutionbanking professionalThe standard for agent identification will determine what checks must be built into the payment flow.fintech companyMultiple parallel verification systems may raise integration costs.Visa, Mastercard and Ant build shared Know-Your-Agent framework as Sumsub links KYC to agent transactionsbanking professionalIdentifying the agent and proving human approval add a new authentication layer to the payment flow.fintech companyJoining shared identification frameworks may become a condition of distribution.American Express publishes its standard for AI agent paymentsfintech companyIf you build an agent payment product and expect card acceptance, align with the AmEx standard.banking professionalPre-payment agent verification offers a model for reshaping fraud screening.Agent identity: LexisNexis, Mastercard and Trulioo build Know Your Agent toolsbanking professionalFor transactions initiated on a customer's behalf by an agent, banks must decide which identification signals they will accept.fintech companyBuilding agent risk signals into the fraud-prevention stack may become an expectation.
Regulators on agent authentication
India prepares AI-agent registry and Unified Agent Protocol for UPIfintech companyUPI agents will only be able to operate in a registered, verified way, which developers need to factor in now.banking professionalBanks will need to verify agent authentication and handle authorised transactions.Fed's Waller: AI shopping changes what banks need to authenticatebanking professionalFraud-detection and authentication models must distinguish between a transaction initiated by the customer and one initiated by their agent.fintech companySpending permissions granted to agents should be designed in tiers, with limits.Bank regulators in the US and Europe tighten scrutiny of AI risksbanking professionalCompile an inventory of AI use, including components of Chinese origin, and have a documented shutdown mechanism for every autonomous agent.chief financial officerExpect stricter supervisory scrutiny of substitution and outsourcing risks in AI systems.Hong Kong becomes first banking regulator to require identity credentials for AI agentsbanking professionalThe agent's identifier must be recorded in the onboarding and risk-management process.OCC's first supervisory guidance on bank AI use rates simple payment agents as high riskbanking professionalInventories, risk classifications and control documentation for AI systems should be prepared for examinations, especially for agents that make payments.fintech companyFintech firms working with banks can expect stricter AI-control questions from partner banks.
Blurring of consent and fraud liability
PSD3 and PSR: paying via AI agents raises new authentication questionsfintech companyAgentic payment flows must be designed around PSR-compatible strong customer authentication.banking professionalFor card and account payments, the handling of authority granted on behalf of an AI agent needs to be clarified.Asian banks must redefine what counts as an authorized payment in the age of AI agentsbanking professionalAuthentication and dispute-resolution rules must be amended for cases where an agent initiated the payment.retail customerCustomers need to understand what authority they give an agent, because it determines their chances of a successful claim later.AI agents challenge payment intent verification and fraud liabilitybanking professionalDevelop intent and consent checks for transactions made by agents acting on a customer's behalf.chief financial officerWhen granting agents payment authority, account for the costs of identity verification, oversight and liability.fintech companyBuild verifiable intent capture and spending limits into your payment product.Survey: 65% of US consumers would blame the AI company if an AI agent commits fraudbanking professionalGiven the customer trust gap, agent features should be introduced in stages, with transparent liability rules.fintech companyCustomers hold the AI company responsible for an agent's actions, so agent errors pose a reputational risk.
Fears for deposits and customer relationships
Meta AI agent triggers selloff in bank, insurer and travel stocks; Insurify blocks Meta's Muse agentinsurerConsumer AI agents may displace or reshape comparison and intermediary channels, so handling agent traffic is a strategic question.fintech companyComparison sites must decide whether to serve third-party AI agents.Apollo economist warns AI agents could trigger bank runs by shifting depositsbanking professionalDeposit-stability and funding-cost models should also be run against automated, yield-seeking behaviour.chief financial officerDeposit pricing and liquidity plans should include AI-driven shifts of funds as a separate scenario.fintech companyHigh-yield accounts may gain prominence through a new channel: agent-based customer acquisition.retail customerAI assistants can automatically move savings to an account with a better rate, which may benefit the customer.Morgan Stanley launches UK savings accounts as US banks fear AI agents could drain cheap depositsretail customerCustomers can move money more quickly to accounts offering higher interest.banking professionalAI agents could weaken deposit loyalty, so banks should prepare for rate competition.chief financial officerDeposit-cost and deposit-retention models should also be tested against agent-driven reallocation scenarios.Hungarian banks and payment players on AI-agent shopping: OTP, SimplePay, Woltbanking professionalIn agent-based shopping, the bank's role and its proximity to the customer are a strategic decision.fintech companyPayment providers need to develop an authorisation model for payments initiated by an agent.Analyses: AI agents could shake bank deposits and payment authenticationbanking professionalDeposit-retention and authentication strategies should be prepared for agent-driven money movement.chief financial officerAgent-driven scenarios for deposit structure and funding cost are worth modelling.
Google search interest
Hungary
| Topic and search term | Weekly | Average |
|---|---|---|
| Banks exclude external agentsbanking security | 0 | |
| Major banks ban external agentsAI agent | 1.7 | |
| Industry agent identification standardsbank customer identification | 0 | |
| Regulators on agent authenticationbank supervision | 0 | |
| Blurring of consent and fraud liabilitybank fraud liability | 0 | |
| Fears for deposits and customer relationshipsdeposit interest rate | 0 |
World
| Topic and search term | Weekly | Average |
|---|---|---|
| Banks exclude external agentsbank account security | 21.9 | |
| Major banks ban external agentsai agents banking | 1.5 | |
| Industry agent identification standardskyc verification | 11.8 | |
| Regulators on agent authenticationbanking supervision | 1.9 | |
| Blurring of consent and fraud liabilitybank fraud liability | 0 | |
| Fears for deposits and customer relationshipssavings account rates | 55.7 |
Google Trends relative search index over the period studied, weekly average. A narrative and its sub-threads appear in a single comparison: 100 is the group's most-searched day, and the figures can only be compared within the group. The search term was chosen by machine analysis; it measures how much the term is searched, not how widespread the narrative is.
Who is affected and how
- retail customer
- Bank of America says the customer may bear responsibility if they hand over bank login credentials to an agent. Only access explicitly authorized by the bank should be used.
- banking professional
- Terms of use and authentication currently assume human access. It is advisable to develop an agent identification and authorization policy before the regulator requires it (see the Hong Kong requirement).
- fintech company
- The agent model that logs in with the customer's credentials is heading for a dead end. The shift must be to bank-authorised access and Know Your Agent identification.
Hungary
There is no news of a Hungarian banking ban during the period under review. At TechShow, leaders from OTP, SimplePay, Wolt, and Decathlon debated who gets closer to the customer if an AI agent buys. Brett King spoke in Budapest about rewriting the payment competition. It is worth watching how Hungarian banks and the regulator handle access granted by customers to external AI assistants.
Weekly trend
Turning points
Meta AI agent triggers selloff in bank, insurer and travel stocks; Insurify blocks Meta's Muse agentThe drop in bank and insurer share prices caused by Meta's Muse agent turned the issue from a technological novelty into a business-model threat.JPMorgan blocks outside AI agents from customer accounts; Schwab also blocksWith JPMorgan's open ban, banks turned against external agents. Within weeks, major US and Canadian banks followed.OCC's first supervisory guidance on bank AI use rates simple payment agents as high riskThe OCC's first supervisory framework classified simple payment agents as high-risk, giving regulatory tailwinds to bank gatekeeping.
Related narratives in the News magazine
Supporting stories
Big banks and big tech clash over regulation of agentic AIbanking professionalThe outcome of the regulatory debate will determine who is liable for transactions initiated by an agent.fintech companyThe bank–tech clash directly affects market-access conditions for agentic services.Visa, Mastercard and Ant seek to align AI agent verification, but disagree on the solutionbanking professionalThe standard for agent identification will determine what checks must be built into the payment flow.fintech companyMultiple parallel verification systems may raise integration costs.Visa, Mastercard and Ant build shared Know-Your-Agent framework as Sumsub links KYC to agent transactionsbanking professionalIdentifying the agent and proving human approval add a new authentication layer to the payment flow.fintech companyJoining shared identification frameworks may become a condition of distribution.Major banks warn that AI shopping bots raise fraud and data-privacy risksretail customerGive card details or payment authority only to a verified AI agent with limited scope, and monitor its transactions.banking professionalCustomer communication and fraud prevention need a separate scenario for unauthorized payments initiated by an agent.Meta AI agent triggers selloff in bank, insurer and travel stocks; Insurify blocks Meta's Muse agentinsurerConsumer AI agents may displace or reshape comparison and intermediary channels, so handling agent traffic is a strategic question.fintech companyComparison sites must decide whether to serve third-party AI agents.Apollo economist warns AI agents could trigger bank runs by shifting depositsbanking professionalDeposit-stability and funding-cost models should also be run against automated, yield-seeking behaviour.chief financial officerDeposit pricing and liquidity plans should include AI-driven shifts of funds as a separate scenario.fintech companyHigh-yield accounts may gain prominence through a new channel: agent-based customer acquisition.retail customerAI assistants can automatically move savings to an account with a better rate, which may benefit the customer.Fed's Waller: AI shopping changes what banks need to authenticatebanking professionalFraud-detection and authentication models must distinguish between a transaction initiated by the customer and one initiated by their agent.fintech companySpending permissions granted to agents should be designed in tiers, with limits.JPMorgan blocks outside AI agents from customer accounts; Schwab also blocksretail customerDo not give an AI agent access to your account: if you do, every step the agent takes may be your responsibility.banking professionalIdentifying and blocking agent traffic, and handling customer consent, should be written into policy.fintech companyA provider offering an outside agent will have no direct entry to large-bank accounts and must negotiate an authorized access route.AI agents challenge payment intent verification and fraud liabilitybanking professionalDevelop intent and consent checks for transactions made by agents acting on a customer's behalf.chief financial officerWhen granting agents payment authority, account for the costs of identity verification, oversight and liability.fintech companyBuild verifiable intent capture and spending limits into your payment product.American Express publishes its standard for AI agent paymentsfintech companyIf you build an agent payment product and expect card acceptance, align with the AmEx standard.banking professionalPre-payment agent verification offers a model for reshaping fraud screening.Self Labs launches Agent Vault as banks lock AI agents out of accountsfintech companyA product category for managing agent access has emerged, responding to banks' bans.US big banks block third-party AI agents from accessing customer accountsretail customerDo not hand your banking login details to an AI agent, because banks may treat such access as prohibited or unauthorised activity.banking professionalRecognising agent traffic and tightening customer authentication rules is a direct operational task.fintech companyAgent-based services must find an authorised API channel instead of the screen- and password-based access that banks prohibit.Morgan Stanley launches UK savings accounts as US banks fear AI agents could drain cheap depositsretail customerCustomers can move money more quickly to accounts offering higher interest.banking professionalAI agents could weaken deposit loyalty, so banks should prepare for rate competition.chief financial officerDeposit-cost and deposit-retention models should also be tested against agent-driven reallocation scenarios.Hungarian banks and payment players on AI-agent shopping: OTP, SimplePay, Woltbanking professionalIn agent-based shopping, the bank's role and its proximity to the customer are a strategic decision.fintech companyPayment providers need to develop an authorisation model for payments initiated by an agent.Major U.S. banks block AI agents that log in with customers' credentialsretail customerBanking passwords and login details should not be given to AI assistants, because the bank prohibits it and the activity is hard to trace.banking professionalBanks already struggle to tell whether the account owner or an AI agent is carrying out an action, so identifying agent traffic is an urgent task.fintech companyAgent products built on sharing customers' credentials rest on a risky business model; official API access is needed.Canada's big banks, including RBC, BMO and TD, block third-party AI agentsretail customerDo not give out banking credentials to use a third-party AI agent for banking tasks, as the bank may block access.banking professionalDetecting and blocking agent access, and communicating with customers, are new tasks for digital channels.fintech companyAgent-based services must expect banks to block access and should seek official integration.JPMorgan, Amazon and Bank of America restrict external AI agents' accessretail customerAccording to the bank, giving an AI agent access to a bank account is at the customer's own risk, so login credentials should not be handed over.banking professionalAgent access needs a dedicated policy and customer communication; general terms of use are not enough.fintech companyAgent-based services may be blocked by large banks, so an official access solution is needed.Survey: 65% of US consumers would blame the AI company if an AI agent commits fraudbanking professionalGiven the customer trust gap, agent features should be introduced in stages, with transparent liability rules.fintech companyCustomers hold the AI company responsible for an agent's actions, so agent errors pose a reputational risk.Analyses: AI agents could shake bank deposits and payment authenticationbanking professionalDeposit-retention and authentication strategies should be prepared for agent-driven money movement.chief financial officerAgent-driven scenarios for deposit structure and funding cost are worth modelling.Hong Kong becomes first banking regulator to require identity credentials for AI agentsbanking professionalThe agent's identifier must be recorded in the onboarding and risk-management process.OCC's first supervisory guidance on bank AI use rates simple payment agents as high riskbanking professionalInventories, risk classifications and control documentation for AI systems should be prepared for examinations, especially for agents that make payments.fintech companyFintech firms working with banks can expect stricter AI-control questions from partner banks.Agent identity: LexisNexis, Mastercard and Trulioo build Know Your Agent toolsbanking professionalFor transactions initiated on a customer's behalf by an agent, banks must decide which identification signals they will accept.fintech companyBuilding agent risk signals into the fraud-prevention stack may become an expectation.
Machine analysis based on story titles and summaries, with a verification pass. The source of the facts is the article linked at each story.