Britain's biggest banks have banned AI agents from accessing customer accounts over fears that personal finance bots will go rogue.
HSBC, Nationwide, NatWest, Barclays and Lloyds have all moved to block AI agents, including in some cases specific bans on making transactions on a customer's behalf without their direct involvement.
AI agents that can manage people's money – and their personal lives – are seen as one of the most promising frontiers for the technology.
However, banks are concerned that bots could go rogue and mismanage people's money, with no clear rules as to who would be responsible in that scenario. Lenders are also concerned that en masse, bot-enabled switches could also threaten profits and, in a worst-case scenario, financial stability.
In September, Mark Zuckerberg's Meta launched Muse , a personal AI assistant that can carry out tasks such as shopping and managing bills for users. The app, which has not yet launched in the UK but is expected to, has topped download charts in North America.
Rather than just providing advice, AI agents such as Muse will carry out tasks such as sending emails or ordering shopping. Increasingly in the US, customers are directing their agents to manage their savings directly in their bank accounts.
Similar products are expected in Britain. Anne Boden, the founder of Starling Bank, has argued that AI should be entrusted with people's finances and recently announced on LinkedIn that she was hiring engineers for a new startup which would be "right at the start of the next revolution in banking technology".
Bank bosses are concerned that bots could go rogue, following a series of incidents outside the financial system in which AI has accessed systems without authorisation.
Manus Costello, the chief financial officer at Standard Chartered, said last week: "We need to be very careful that those tools are safe for customers and safe for us before we allow them in."
Will Chalmers, the chief financial officer of Lloyds, said recently the bank had not seen any sign of AI agents accessing bank accounts but added: "We do guard against that type of risk, and we take it very seriously."
The Financial Conduct Authority is working with the City and other regulators to develop rules to govern AI agents in finance.
Sources close to the regulator said it was up to the banks to make decisions on AI access but added that early research showed that agents could behave in unexpected ways.
The bans from UK banks are understood to be pre-emptive, with no signs yet that AI bots are accessing accounts.
As well as concerns about rogue AI, the technology also poses a potential threat to banks.
McKinsey has warned that savers could shift up to £700bn out of accounts that offer little or no returns and into savings products on the advice of AI chatbots.
If just 5pc to 10pc of bank balances are moved this way, the profits lenders make on deposits could drop by a fifth or more.
Anthropic and OpenAI both recently launched personal finance products that plug in to people's accounts and give them advice on what to do with their money. However, the services only read financial information, rather than taking action on users' behalf.
'You don't want rogue agents'
There are also fears that AI agents acting in unison could pull deposits from a bank at the same moment, triggering a run at machine speed.
Jamie Dimon, the boss of JP Morgan, warned earlier this year : "It's a new world and with all these bots and agents, could someone ... cause a problem? Maybe."
The bans by British banks follow similar moves in the US and Canada, where Citigroup, Bank of America, Wells Fargo and JP Morgan have all blocked AI agents from accessing accounts.
"You don't want rogue agents, and there could be thousands of them," Mr Dimon said during the bank's third-quarter earnings call.
Rick Wurster, the chief executive of Charles Schwab, the US brokerage giant, said that the company was actively blocking between 10 and 20 clients a week from using AI agents.
"We don't want someone just giving their password to Muse, and letting Muse come in and take action on that client's behalf," Mr Wurster said in a recent interview. "We can't control it, the client can't control it, something could go wrong and who's liable?"
AI agents face tougher guardrails in the UK, where strict authentication rules mean most payments must be approved through a mobile banking app, often with a fingerprint or face scan.
