UK consumers trust artificial intelligence more with their health and their legal problems than they do with their money, according to new research from Enfuce, Europe's card issuing and processing powerhouse. The findings point to a payments-specific trust gap. And as AI moves deeper behind every transaction, it's arriving in the one place people already struggle to see, understand and trust.Key findings:37% of UK adults would trust AI to help with a health symptom and 38% with a legal question, and just 22% for making or managing a paymentOf ten everyday tasks tested, payments ranked last for AI trust, behind general search (63%), navigation (57%) and recommendations (56%), and last for actual use: among AI users, 53% have used it for general search and 26% for health symptoms, versus 10% for savings and investments and 8% for paymentsConsumers also find payments hard to understand: 48% think only one or two companies are involved in processing a card payment, and just 27% feel they understand well how their payments are kept secure45% would feel uncomfortable making a payment if they knew fraud detection was handled by AI rather than a human team; only 23% would feel comfortablePayments already have an accountability blind spotThe reluctance to trust AI with money reflects a payments system that is not well understood and, as a result, not fully trusted. Almost half (48%) think only one or two companies are involved in processing a card payment, when a typical transaction can involve five or more, and just 27% feel they understand well how their payments are kept secure.Without visibility into the system, confidence rests on who people recognise: 88% trust banks and card networks to keep their money secure, compared to just 41% for fintech and payment apps. The same logic shapes which methods feel safe, the ones consumers know and understand rate highest. Cash and direct debit both score 82%, while newer, less visible ways to pay inspire less confidence: contactless at 64% and digital wallets at 57%.The system also has an accountability problem when things go wrong. No single party is named as mainly responsible for the loss by more than a quarter of consumers, 23% point to the platform where the scam originated, 21% to the consumer, 15% to the consumer's bank, and 15% simply don't know.Asked what would most increase their trust in a payment method, 30% said knowing they would get their money back if something goes wrong, more than the next three factors combined: strong fraud protection (19%), personal data protection (8%) and provider reputation (6%).Even as a safeguard, AI isn't trustedConsumers are reluctant to involve AI in payments even when its role is purely protective. 95% would accept at least one additional safeguard in exchange for more secure payments, including approval being required before a payment (47%), additional identity checks (43%), reimbursement protection and access to human support (36%). Yet just 19% would accept AI-led fraud detection as that safeguard and 45% would feel uncomfortable making a payment knowing fraud detection was handled by AI rather than a human team, against only 23% who would feel comfortable.Nine in ten (90%) have at least one concern about AI being used to detect payment fraud, led by fear of AI making mistakes (51%), data privacy (39%) and cybersecurity risk (35%). Even in a protective role, AI has yet to earn the trust consumers place in human oversight.Taken together, the findings point to a payments system where trust rests on what consumers can see and understand, and where AI, for now, sits outside that. “We spent decades making payments invisible, tap, pay, move on. Yet our research shows people are more willing to trust AI with a legal question than with a payment. Our industry needs to take that hesitation seriously. Customers should be able to trust that their money is protected, that someone is accountable for it, and that there is somewhere to turn when something goes wrong. Payments can stay invisible. Trust cannot.” Denise Johansson, CEO and Co-founder, Enfuce.NoYesArtificial Intelligence01 Oct, 2026