At the IMF and World Bank meetings in Bangkok, the heads of the Bank of England and the IMF warned that fraud, sped up by AI, could make people stop trusting banks. The IMF's chief described growing up behind the iron curtain with her savings in a shoebox under the bed, and warned that she fears a return to that. Digital fraud already costs the world between $500 billion and $1 trillion a year, by the usual estimates, and the meetings launched a set of principles for countries called the Bangkok blueprint.
Cheap fraud cannot be fought by asking customers to be careful
A scammer who uses AI can write a believable message in any language and hold thousands of conversations at once. That pushes the cost of each attempt close to zero, so even a very careful person will meet enough attempts that one will get through. A defence that depends on customers spotting every trick fails at that volume. What works is stopping the payment, and only the bank, the platform or the phone company can do that. So the question that matters is which of them is made to pay.
Britain made banks pay, and the banks started watching each other
Since October 2024, a British bank that sends a scam payment and the bank that receives it split the cost of repaying the victim equally. That detail changes what each bank does. The sending bank questions unusual payments, and the receiving bank now has a reason to find and close the accounts that scammers use to collect money, because each of them is paying half.
The rule has a limit that shows the same logic from the other side. It does not cover international payments, and losses on those scams rose from £21 million to £60 million between 2023 and 2025. Where no one is made to pay, the money keeps leaking.
Platforms are the payer still missing
Take a homeowner in Manchester who sees an advert on social media for a cheap boiler service and pays £3,000 by bank transfer. The boiler firm does not exist. Under British rules the bank repays her. The platform was paid to show the advert, pays nothing for the loss, and has little reason to check the next advertiser harder.
The governor's complaint is about exactly this. Newspapers answer in law for what they print, and platforms mostly do not. The EU's agreed rules would let a bank recover its refunds from a platform that left fraudulent content up after being told about it, and would make financial advertisers prove they are authorised. Those EU rules still need formal adoption. Australia's law, with most duties starting in March 2027, covers banks, phone companies and platforms together.
My expectation is that most large markets will make platforms pay part of the bill within a few years. When they do, the checks on adverts and on new accounts will get stricter, and honest small businesses that advertise will face more identity verification before their first campaign runs. That is a cost, but a smaller one than customers losing trust in every payment link they are sent.
The governor also argued that the same technology that arms the attackers can arm the defence. It will, and the buyers will be whoever has been made to pay: banks in Britain now, and platforms once the law reaches them. Until it does, how well your money is protected from a scam depends less on how careful you are than on which country you live in and who its law makes responsible.