Banks Could Face Bans on Dangerous AI After This Research

Jan de Vries · 2026-08-13

Artificial intelligence is transforming the banking world at breakneck speed. But a new study suggests that some of these tools might be doing more harm than good, and regulators are starting to take notice. The push for tougher safeguards comes from a surprising corner: Durham University. Their researchers have been digging into how banks use AI, and what they found has sparked a serious conversation about whether certain algorithms should be allowed to operate at all. ### The Core Problem With AI in Banking Here's the thing: banks are using AI for everything from approving loans to detecting fraud. On the surface, that sounds great. Faster decisions, lower costs, and maybe even fewer errors. But the Durham University research highlights a darker side that often gets overlooked. Some AI systems are what experts call "black boxes." They make decisions, but nobody can really explain how they got there. That's a huge problem when a bank denies your mortgage application or freezes your account. If the AI can't explain itself, how do you challenge its decision? The research also points to something even more concerning. Certain AI models can inadvertently pick up biases from the data they're trained on. That means they might discriminate against specific groups of people without anyone ever programming them to do so. It's not intentional, but the damage is very real. ### Why "Dangerous" AI Is Different Not all AI is created equal, and the researchers are careful to make that distinction. Some algorithms are relatively simple and transparent. You can trace every step they take. Those are generally fine. But the dangerous ones are different. They're the ones that make high-stakes decisions with little to no human oversight. Think about a system that automatically decides who gets credit and who doesn't, or one that flags certain transactions as fraudulent. When these systems get it wrong, real people suffer real consequences. The Durham University team isn't saying we should ditch AI altogether. That would be throwing the baby out with the bathwater. Instead, they're calling for a more measured approach, one that puts safety and fairness ahead of pure efficiency. ### What a Ban Could Look Like So what would an actual ban entail? It's not as simple as flipping a switch and turning everything off. The proposal is more nuanced than that. Regulators might start by identifying specific types of AI that are deemed too risky for use in financial services. Those would be the ones that get banned or heavily restricted. Other, safer applications could continue to operate, but under stricter oversight. - Banks would likely need to prove their AI systems are explainable - Regular audits would become mandatory, not optional - Human review would be required for any high-stakes decision - Fines for non-compliance could be substantial The goal isn't to stop innovation. It's to make sure innovation doesn't come at the expense of consumer protection. ### The Broader Implications for Financial Services This isn't just a European issue, and that's why it matters to professionals in the United States. Financial markets are deeply interconnected. What happens in London or Frankfurt often ripples across the Atlantic. If European regulators move forward with bans or stricter rules, American banks might face pressure to follow suit. After all, a global bank operating in both markets can't easily run two completely different AI strategies. There's also the question of public trust. When people hear that AI could be making unfair decisions about their money, they start to get nervous. That nervousness can spread quickly, and banks know it. ### What Happens Next The conversation is just getting started. Durham University's research has opened the door, and now it's up to regulators, bankers, and technologists to figure out where to go from here. For now, the message is clear: AI in banking needs guardrails. Not because AI is inherently bad, but because it's powerful, and power without oversight is a recipe for trouble. Whether we'll see actual bans remains to be seen. But one thing's for sure: the days of letting algorithms run wild in financial services are probably numbered. That's a change worth watching, no matter which side of the Atlantic you're on.