- According to The Telegraph on 8 November 2026, major UK banks, including HSBC, Nationwide, NatWest, Barclays and Lloyds, have introduced measures to prevent AI agents from accessing customer accounts, particularly when transactions are carried out without the account holder’s direct involvement.
- The main concern is that AI agents could execute incorrect transactions, exceed their authorised permissions or behave unpredictably. Meanwhile, legal responsibility among banks, customers and AI providers remains unclear.
- Banks are also concerned that AI agents could automatically move deposits into higher-interest products, reducing their deposit funding and profits from traditional savings accounts.
- McKinsey warns that AI could trigger the movement of up to £700 billion away from low-interest or non-interest-bearing accounts. If just 5–10% of deposit balances were transferred, banks’ profits from deposit activities could fall by 20% or more.
- A more serious risk is that multiple AI agents could withdraw funds from the same bank simultaneously, triggering a bank run at machine speed, significantly faster than ordinary customer reactions.
- In the US and Canada, Citigroup, Bank of America, Wells Fargo and JPMorgan have also introduced similar restrictions. Charles Schwab says it prevents around 10–20 customers each week from using AI agents to access their accounts.
- JPMorgan CEO Jamie Dimon has warned that thousands of AI agents operating beyond human control could create new risks for the financial system. The UK’s Financial Conduct Authority (FCA) is working with stakeholders to develop rules governing AI agents in banking.
- Unlike AI agents with transaction authority, Claude Personal Finance and ChatGPT Finances currently have read-only access to financial data and cannot directly transfer money or modify accounts.
📌 UK banks’ decision to block AI agents highlights the important boundary between AI that provides financial advice and AI that is authorised to execute transactions. Beyond individual errors, AI agents could also create systemic risks if thousands of automated agents react simultaneously. Notably, the threat is not limited to technological security but also extends to banks’ business models. Regulating financial AI agents requires a combination of transaction-permission limits, customer authentication, clear accountability and controls over collective behaviour.
