Banks want AI to manage your money. First ask who pays when it goes wrong, writes Dean Dunham KC
Innovation should make consumers’ lives easier, not leave them facing a legal nightmare
Imagine waking up tomorrow morning, checking your bank account and discovering that £20,000 has disappeared.
You haven’t been scammed. You haven’t given your banking details to a criminal and nobody has hacked your account.
Instead, the money vanished because an artificial intelligence assistant you authorised to manage your finances made a catastrophic mistake.
You contact your bank, expecting an immediate refund, but the bank says the transaction was authorised because you gave the AI permission to act on your behalf.
You contact the technology company, which points to its terms and conditions and argues it cannot be held responsible for every decision its software makes.
Suddenly you’re caught in the middle, thousands of pounds out of pocket, with nobody willing to accept responsibility.
Welcome to what could become the next major consumer protection scandal.
This is not a far-fetched concern.
Andrew Bailey, Governor of the Bank of England, has warned about the legal and regulatory challenges created by increasingly autonomous AI systems that can make financial decisions and carry out transactions on behalf of consumers.
I believe he is right to be concerned.
Artificial intelligence is moving at an extraordinary pace. We are rapidly approaching a world where consumers could routinely use AI assistants to compare financial products, move savings, pay bills, make purchases and manage their money.
The potential benefits are enormous.
Imagine a digital assistant that constantly searches for better savings rates, identifies unnecessary subscriptions, challenges excessive charges and makes sure you never pay more than necessary for everyday services.
For consumers, that could be revolutionary.
But there is a fundamental problem. What happens when the technology gets it wrong?
As a consumer lawyer and broadcaster, I have spent years explaining the protections available when banks make mistakes, consumers fall victim to fraud or financial institutions fail to meet their legal obligations.
Those protections matter because consumers are often at a significant disadvantage when dealing with powerful financial institutions.
But AI introduces an entirely new complication.
Traditionally, there is a distinction between a transaction you authorise and one you do not.
That distinction is central to determining when banks must reimburse customers.
Now imagine you authorise an AI assistant to transfer £5,000 into a savings account offering a better interest rate.
Instead, the AI transfers £50,000.
Or it sends the money to an incorrect recipient.
Or, worse still, a fraudster manipulates it into transferring your savings into a criminal’s account.
You authorised the AI to act for you, but you certainly didn’t authorise those particular outcomes.
So where does responsibility lie?
Under existing UK payment services law, consumers have important protections against unauthorised transactions. But whether a payment initiated by an AI assistant is legally authorised could depend on the permissions given, how the payment was authenticated and precisely what went wrong.
It would be wrong to suggest that banks can automatically escape liability simply because AI was involved.
Equally, it would be wrong to assume every mistake made by an AI assistant must result in an automatic refund and that uncertainty is precisely why this issue needs addressing now.
Consumers should not be expected to understand complicated questions of artificial intelligence, payment authorisation and contractual liability simply to recover their own money.
Nor should banks and technology companies be allowed to use complicated contractual arrangements to leave customers caught in the middle.
There is another issue that concerns me.
Banks and technology companies are already competing for control over how consumers access and manage their finances.
AI assistants could eventually become incredibly powerful tools for consumers.
Think about your savings.
How many people leave thousands of pounds sitting in accounts at miserable interest rates because they don’t have time to search for better deals?
An AI assistant could potentially solve that problem by constantly identifying better offers and, with appropriate permission, moving money to maximise returns.
That sounds fantastic for consumers.
But perhaps not so fantastic for banks that benefit from customers leaving their money in poorly paying accounts.
When banks restrict AI assistants from accessing customer accounts, are they protecting customers from security risks, protecting their own commercial interests, or doing a bit of both?
There are legitimate security concerns. Giving third-party technology access to financial accounts creates obvious risks, particularly if the systems can initiate transactions without constant human supervision.
But consumers must not lose out on innovation simply because it threatens established business models.
The solution is not to prevent AI from transforming financial services.
The solution is to make sure consumer protection keeps pace.
I believe we need clear legal rules establishing who is responsible when an AI system makes a financial mistake.
Banks should remain accountable for their own failures and for meeting their existing payment services obligations.
Technology providers should be accountable where their products are defective, insecure or fail to operate as promised.
And consumers should be given clear information about exactly what authority they are granting an AI assistant, including any financial limits and the risks involved.
There should also be straightforward ways to suspend an AI assistant’s access, challenge transactions and resolve disputes without being passed endlessly between different companies.
Most importantly, the Government and financial regulators need to act before widespread consumer harm occurs, rather than waiting until thousands of people have lost money.
We have seen what happens when technology moves faster than consumer protection.
Online scams have become increasingly sophisticated. Fraudsters have exploited weaknesses in payment systems and consumers have too often been left fighting lengthy battles to recover money they should never have lost.
We cannot afford to repeat those mistakes with artificial intelligence.
Because here’s the reality.
AI might be intelligent enough to manage your bank account, but it cannot stand before a court and personally reimburse you when something goes wrong.
Ultimately, there must be a legally accountable person or business behind the technology and that responsibility must be clear from the outset.
My message to the banks, technology companies and regulators is simple: if you want consumers to trust AI with their money, you must first make absolutely clear who picks up the bill when it makes a mistake.
Innovation should make consumers’ lives easier, not leave them facing a legal nightmare.
