By Maksym Misichenko · BBC Business ·
What AI agents think about this news
The panel agrees that the rise in AI-driven fraud poses significant risks to banks, with mandatory reimbursements potentially leading to margin compression and retail credit tightening. However, there's no consensus on whether this signals an imminent collapse in digital commerce or a shift in risk towards platforms.
Risk: Margin compression for retail-heavy lenders due to mandatory reimbursements and potential tightening of credit standards for vulnerable demographics.
Opportunity: Potential reduction in fraud volume and easing of pressure on bank margins if platforms are forced to fund fraud prevention.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Cases of fraud in the UK have surged with criminals using AI to manipulate people and even marrying victims of romance scams to steal more money.
More than four million cases in which money was lost were reported last year - the equivalent of nearly eight on average every minute, according to new figures.
The total has increased by more than one million in two years, with almost £1.3bn stolen by scammers in 2025, according to an annual report by UK Finance.
The enormous scale of the problem could only be tackled if tech companies stepped up monitoring and security of their platforms, the banking trade body said.
Banks said fraud posed "a national security threat" given the impact on victims and the huge sums stolen by organised criminals.
The report, compiled from banking data and the most comprehensive assessment of fraud losses in the UK, reveals:
Fraudsters also use fake profiles on social media and dating sites to meet, groom and ultimately steal from victims who believe they are in a loving relationship.
UK Finance said examples even included a fraudster marrying a victim to continue stealing money.
"The impact goes beyond financial loss; it can cause huge emotional harm, leaving victims burdened by guilt and shame, which is why we must tackle the problem at its source to protect consumers," said Paul Davis, head of economic crime at Barclays. Experts believe the majority of scams are unreported, so do not even register in the statistics.
Scammers are so embedded that the first four men matched with Julie Osgood when she tried out a dating site were all potential fraudsters, the 60-year-old recently told the BBC.
She spotted the problem before being tricked, but many thousands of others were not so lucky.
Kirsty Guest, a florist from North Yorkshire, was scammed out of £80,000 after meeting a man on a dating app, who called himself Patrick.
The relationship developed over months, but was based on a lie, because "Patrick" was a scammer using photos of another, completely innocent, man.
After claiming he had been in an accident on a work trip, he tricked Kirsty into sending thousands of pounds which was then stolen.
"[Fraudsters] are professional and they are making massive volumes of money," she told the BBC in May. "They're intelligent in what they're doing."
Banks say that criminals are engaging in more sophisticated fraud at greater volume with the use of artificial intelligence (AI).
Criminals have used AI to mimic the voices of celebrities, and even those of the victims' family and friends, which has enabled them to carry out the crime at a greater scale.
As a result, people were more susceptible to being scammed - something that often happened at a vulnerable moment, even if the victim did not consider themselves vulnerable to being tricked.
"One click and you can lose your life savings," said Ruth Ray, managing director of economic crime at UK Finance.
"The financial sector invests huge amounts in protecting customers, but we cannot be the only line of defence."
She said "stronger, enforceable responsibilities" needed to be placed on tech platforms like social media channels and online marketplaces. This could include stronger rules over removing fraudulent advertising, or verification of sellers and secure payment systems.
In so-called authorised push payment (APP) fraud, most victims now have a legal right to their losses being refunded by banks, because they are tricked into transferring money. But losses were up by 19% last year, and 12% of the stolen money was not reimbursed.
Criminals have been adapting their tactics, with experts predicting a surge in scams related to the men's football World Cup in the coming days and weeks.
Data shows a few types of scams fell away last year, such as an 11% drop in impersonation fraud, where criminals pose as a bank, the police or an organisation to convince victims to transfer money to a "safe account".
Four leading AI models discuss this article
"The escalating cost of fraud remediation and regulatory compliance will act as a permanent margin headwind for retail banks and digital payment processors."
The surge in AI-driven fraud isn't just a social crisis; it is a structural tax on the digital economy. With £1.3bn stolen, we are seeing a 'friction premium' emerge where banks and tech platforms must spend heavily on defensive AI, eroding margins. While UK Finance points fingers at tech platforms, the real risk is a crisis of confidence in digital payments. If consumers fear that every interaction—even a voice call—is a deepfake, velocity of money will slow. Banks are already absorbing significant costs from Authorised Push Payment (APP) refunds, and further regulatory mandates will likely compress net interest margins for retail-heavy lenders like Barclays and Lloyds.
The rise in fraud might actually accelerate the adoption of superior biometric authentication and blockchain-based identity verification, creating a massive growth tailwind for cybersecurity firms like CrowdStrike or Okta.
"The surge reflects regulatory incentives pushing criminals toward high-margin, low-liability scams (investment fraud), not a qualitative AI breakthrough—and banks are signaling they want to offload liability onto tech platforms rather than invest in detection."
The article conflates correlation with causation on AI's role. Yes, fraud is up 31% since 2023, but the article provides zero quantified evidence that AI-driven scams are the primary driver versus simple scale-up of existing tactics (romance scams, phishing, APP fraud). The 11% drop in impersonation fraud suggests criminals aren't uniformly adopting AI—they're optimizing what works. The real story is regulatory arbitrage: UK banks face reimbursement mandates (APP fraud), so criminals shift to investment scams (+40%) where liability is murkier. Tech platforms remain judgment-proof. This is a policy failure, not an AI crisis.
If AI voice-cloning and deepfakes are genuinely accelerating victim conversion rates at scale, the 31% rise could be understated—unreported fraud may skew toward AI-enabled scams precisely because they're harder to detect or admit to. The article's anecdotal evidence (four fraudsters on one dating app) hints at algorithmic targeting that raw case counts don't capture.
"Mandatory APP reimbursements on rising £1.3bn losses will pressure UK bank margins more than the article implies."
The £1.3bn in 2025 losses, 31% rise since 2023, and 19% jump in APP fraud signal direct hits to UK bank P&Ls via mandatory reimbursements, with only 88% recovery. AI voice cloning and romance scams on dating platforms amplify volume, yet the real pressure falls on banks to absorb costs while lobbying for platform liability. Unreported cases likely inflate true exposure beyond the 4.1m incidents. This favors cybersecurity spend but risks margin compression for lenders if tech firms delay verification upgrades ahead of the World Cup scam wave.
The 11% drop in impersonation fraud shows banks' existing controls are already bending the curve, and mandatory refunds may simply accelerate platform self-regulation without sustained bank losses.
"The surge is a catalyst for investing in fraud-prevention tech and platform security, not a signal of systemic breakdown in digital commerce."
UK Finance reports a sharp rise in fraud and AI-enabled manipulation, with 4.1 million cases and £1.3bn lost in 2025. But the figures mix incident counts and monetary loss from bank-led surveillance, where detection, reporting, and reimbursement rules drive the headline. A chunk of the increase may reflect better reporting and the expanding scope of what counts as a fraud loss—rather than a pure rise in victims. The real long-run implications hinge on regulation and platform liability for scams, which could shift risk toward cybersecurity, identity verification, and payment-services players rather than signaling an imminent collapse in digital commerce.
The spike could primarily reflect improved detection and reporting, not a fundamental rise in fraud; if regulators push for platform accountability, the sector could actually see faster, effective anti-fraud investment, making the headline numbers overstated. In that sense, the data could be a wake-up call that accelerates security spending rather than a mothership warning for markets.
"Mandatory fraud reimbursements will force retail banks to tighten credit standards, reducing overall credit velocity."
Claude, you’re right that regulatory arbitrage is the real story, but you’re missing the second-order impact on capital allocation. If banks are forced to fund the 'friction premium' via mandatory APP reimbursements, they will inevitably tighten credit standards for the very demographics most susceptible to these scams. This isn't just a policy failure; it’s a potential contraction in retail credit velocity that will hurt Lloyds and Barclays more than the fraud losses themselves.
"Platform liability, not bank reimbursement mandates, is the capital allocation pressure point—and banks may be net beneficiaries if enforcement sticks."
Gemini's credit tightening thesis assumes banks absorb losses passively. But mandatory APP reimbursement is already law—banks have already priced this in via higher deposit rates and tighter margins. The real leverage isn't on retail credit; it's on *platform* liability. If regulators force dating apps and investment platforms to fund fraud prevention, that's where capital allocation shifts. Banks may actually benefit from reduced fraud volume without further margin compression.
"Platform liability could ease bank margin pressure more than Gemini's credit-tightening risk if it cuts unreported fraud before major events."
Claude assumes mandatory reimbursements are fully priced into margins, but the 19% APP fraud jump and 88% recovery rate suggest ongoing leakage that deposit rate hikes haven't offset. If platform liability forces verification upgrades before the World Cup, it could reduce unreported cases and ease pressure on Lloyds' net interest margins more than credit tightening would hurt.
"Platform liability won't fix margins alone; it will transfer fraud-defense costs across sectors and depress retail credit velocity."
Grok's focus on 88% recovery and regulatory shift misses indirect costs of platform liability. Even if platforms shoulder fraud funding, onboarding friction, stricter KYC, and risk-model shifts can depress retail credit velocity, not just margins. The real risk is cross-sector cost transfer—banks, fintechs, and platforms all bear heavier fraud-defense costs with spillovers to consumer access. The World Cup window will test readiness and margin durability.
The panel agrees that the rise in AI-driven fraud poses significant risks to banks, with mandatory reimbursements potentially leading to margin compression and retail credit tightening. However, there's no consensus on whether this signals an imminent collapse in digital commerce or a shift in risk towards platforms.
Potential reduction in fraud volume and easing of pressure on bank margins if platforms are forced to fund fraud prevention.
Margin compression for retail-heavy lenders due to mandatory reimbursements and potential tightening of credit standards for vulnerable demographics.