# Five large US banks move to block customers' AI agents as a new benchmark shows end-state scoring overstates agent success

JPMorgan Chase, Morgan Stanley, Bank of America, Citi and Wells Fargo cannot tell agent sessions from human ones on shared credentials, and an Indian payments-network benchmark finds that up to 12 percent of "successful" agent sessions skipped required steps.

- Published: 2026-10-24T06:20:43.691Z
- Canonical: https://polylog.news/ai/2026-10-24/five-large-us-banks-move-to-block-customers-ai-agents-as-a-n
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news), [Axios](https://www.axios.com/2026/10/22/ai-agents-banks-unauthorized-accounts-access), [arXiv cs.CL](https://arxiv.org/abs/2610.30161), [Tom's Hardware](https://www.tomshardware.com/tech-industry/cyber-security/us-banks-scramble-to-restrict-ai-agent-activity-agents-detected-accessing-customer-accounts-or-even-attempting-to-make-transactions-in-breach-of-policy)

According to [Axios](https://www.axios.com/2026/10/22/ai-agents-banks-unauthorized-accounts-access), JPMorgan Chase, Morgan Stanley, Bank of America, Wells Fargo and Citi are each limiting third-party agent access to customer accounts. Customers had given their own login credentials to autonomous assistants, which then [carried out online-banking operations on their behalf](https://t.me/ai_machinelearning_big_data/11196). The banks can see which credentials logged in. They often cannot tell whether a human or an agent directed the session, or whether the customer authorized a specific action. Some banks [detected agents attempting transactions that breached policy](https://www.tomshardware.com/tech-industry/cyber-security/us-banks-scramble-to-restrict-ai-agent-activity-agents-detected-accessing-customer-accounts-or-even-attempting-to-make-transactions-in-breach-of-policy).

JPMorgan allows limited, view-only data retrieval through aggregators such as Plaid. That route does not let agents move funds or pay bills. Jamie Dimon, JPMorgan's chief executive, told investors the bank would not grant "just any" outside agent broad access to its systems.

New research on Indian retail banking shows why procedure matters as much as outcome. Krishanu Adhikary and Prashant Devadiga of the National Payments Corporation of India (NPCI) published [TauIndianBankBench](https://arxiv.org/abs/2610.30161), a set of 1,000 tasks in which an assistant serves a simulated customer under bank rules. An example rule is that the assistant must obtain a closure quote before it closes a deposit. Many required steps leave the final database state unchanged. The benchmark therefore checks that the agent took every required step and made no unrequested change. Across the models tested, [6.5 to 12.0 percent of sessions](https://arxiv.org/html/2610.30161) passed an end-state-only check even though the agent skipped a required step or made an unrequested change. Tau-style benchmarks that grade only final records would count those sessions as successes.

## What this means

Consumer agents that operate a browser on a user's credentials depend on incumbent platforms tolerating them, and the largest US banks have now said they will not. The losers are browser-use agent products that rely on credential sharing for financial tasks. The winners are aggregators with sanctioned, scoped APIs (the Plaid model) and any vendor that can offer delegated, auditable agent authorization. Banks want control over liability and fraud exposure. They can enforce that through terms of service and bot detection, which makes it a binding constraint. The NPCI result supplies the technical reason banks are right to worry: end-state grading misses roughly one in ten procedural failures. Open Banking's API mandates and the end of screen-scraping in the UK and EU are the precedent. Platforms pushed access onto authorized, scoped interfaces. The likelier outcome is that, within months, banks publish agent-specific API terms with read and write scopes and per-action consent, and agent vendors move to them. The condition is that regulators do not order banks to honor credential-based agent access. A Consumer Financial Protection Bureau move that treats agent access as a customer data right would reverse this.

## What to watch

- Whether any of the five banks publishes a formal agent-access API or scope model. That would show the restriction is a step toward paid, controlled access and not a permanent ban.
- How the Consumer Financial Protection Bureau's open-banking rules are applied to AI agents. They decide whether customers have a legal right to delegate account access to software.
- Whether agent benchmark maintainers adopt procedure-level grading like NPCI's. Leaderboard scores on banking and support tasks would fall if they did.