Five major UK banks restrict AI agents’ access to customer accounts
The Telegraph reported on November 8 that HSBC, Barclays, Lloyds, NatWest and Nationwide have taken measures to block or restrict AI agents’ access to customer accounts. Some banks explicitly prohibit agents from carrying out transactions on behalf of customers when the customers are not directly involved.
Lloyds Chief Financial Officer Will Chalmers said the bank has not yet found signs of AI agents actually accessing accounts, but it has put safeguards in place against related risks. The Financial Conduct Authority is working with the industry and other regulators to develop rules for AI agents in finance. The report said these restrictions are currently precautionary measures, and there is no evidence yet that AI agents have entered UK bank accounts.
McKinsey previously analyzed that AI agents could prompt customers to move funds from low-yield or non-interest-bearing accounts into higher-yield products, with a potential scale of up to £700 billion. Banks are concerned that large-scale fund transfers could affect net interest margin profits and, in extreme cases, pose risks to financial stability. There are currently no clear rules specifying which party would be responsible if an AI agent’s misoperation caused losses.
Original source:
Title: Banks ban AI from accounts over fears of rogue agents Publisher: Tom Saunders / The Telegraph Publication date: 2026-11-08 URL: https://www.telegraph.co.uk/business/2026/11/08/banks-ban-ai-from-accounts-over-fears-of-rogue-agents/ Nature of source: media report Verification status: full text read in this round, verification date 2026-11-08, publicly accessible Supporting content: supports the five banks restricting AI agents’ access to accounts, precautionary measures, unclear allocation of responsibility, and McKinsey’s statement about the potential scale of fund transfers
ABAB AI Insight
The restrictions by UK banks target the path for third-party AI agents to log in directly or execute transactions on behalf of customers. Lloyds said it has not observed actual access, indicating that the current measures are not aimed at incidents that have already occurred, but at scenarios that have not yet appeared but may appear. Compared with similar restrictions taken in October 2026 by U.S. institutions such as JPMorgan Chase, Citi and Bank of America, the actions by UK banks came later than their U.S. peers, but the scope is likewise concentrated in large retail banks.
In reports from September 2026 and earlier, McKinsey said that if AI agents help customers shift 5% to 10% of low-yield deposits into higher-yield products, deposit-related profits could decline by about 20%. This calculation is based on the reality that customers currently hold large amounts of low-interest or non-interest-bearing demand deposits. Restricting agents from direct operation is equivalent to adding a manual confirmation step between customer-side AI tools and account operations, thereby reducing the likelihood of large-scale automated transfers.
Allocation of responsibility is the key unresolved issue. There are currently no clear regulations specifying which party would bear compensation liability if an AI agent misoperates. The Financial Conduct Authority is discussing rules with the industry, but no mandatory standard has yet been formed. Banks chose to impose restrictions first, drawing their own boundaries before the regulatory framework is complete. Unlike banks’ internal use of AI for customer service or fraud detection, the object of this restriction is customer-authorized external agents.
This event changes banks’ access permissions for customer-side AI agents, rather than the scope of banks’ own AI applications. The indicators to watch are whether the FCA will issue guidance on liability for AI agent transactions in the coming months, and whether any bank will launch certified read-only or manual-confirmation interfaces after the restrictions.
ABAB News · Cognitive Law
Precautionary restrictions occur before an incident, and when responsibility rules are lacking, banks prioritize closing direct-operation paths.
The scale of automatic fund transfers depends on whether customer-side tools can bypass the confirmation threshold set by banks.