Banks Could Face Bans on Dangerous AI After New Research

Jan de Vries · 2026-08-13

Artificial intelligence is reshaping the financial world, but not everyone is convinced it's for the better. A new study from Durham University is turning heads by calling for tougher safeguards on AI in banking, and some experts say outright bans might be on the table. If you work in finance, fintech, or even just care about where your money goes, this is a conversation worth paying attention to. The core issue isn't AI itself, which has been helping banks detect fraud and streamline customer service for years. The problem is the kind of AI that makes decisions with little to no human oversight, especially when those decisions affect your ability to get a loan, open an account, or access your own funds. Durham's researchers argue that some of these systems are simply too risky to deploy without stricter controls. ### What the Durham University Research Actually Found The team looked at how banks currently use AI across their operations, from credit scoring to real-time trading algorithms. What they found was a patchwork of rules that varies wildly from one institution to another. Some banks have robust internal review boards, while others rely on off-the-shelf software that no one fully understands. That lack of transparency is what the researchers are most worried about. When an AI system denies a mortgage application or freezes a card, the customer rarely gets a clear explanation. And when the system makes a mistake, there's often no clear path to appeal. The study suggests that without meaningful oversight, these tools could cause real financial harm. The report stops short of calling for a complete industry shutdown, but it does recommend a set of stricter standards. These include mandatory stress tests for high-risk AI models, clear accountability for any decisions made by algorithms, and a requirement that banks prove their systems are fair before rolling them out. ### Why a Ban Might Be on the Table The word "ban" sounds extreme, but it's not as far-fetched as you might think. Regulators in the United States have already shown they're willing to step in when technology outpaces the rules. The Consumer Financial Protection Bureau has been vocal about the dangers of black-box algorithms, and the Federal Trade Commission has fined companies for using biased AI. In Europe, the conversation is even further along. The EU's AI Act, which is currently being phased in, classifies certain uses of AI as "high-risk" and imposes heavy penalties for non-compliance. Durham's research aligns with that momentum, suggesting that similar logic should apply to the financial sector on a global scale. A ban wouldn't necessarily mean ripping out every AI system in banking. More likely, it would target the most dangerous applications, like predictive algorithms that decide who gets credit based on opaque criteria. In those cases, a temporary halt could give regulators time to figure out what safe deployment actually looks like. ### What This Means for You If you're a consumer, this news is mostly reassuring. It means the people in charge are finally asking hard questions about the tools that manage your money. But if you work in fintech or banking operations, it's a heads-up that the ground is shifting under your feet. Compliance teams should start auditing their AI models now, not later. That means documenting every decision the algorithm makes, testing for bias on a regular basis, and keeping humans in the loop for anything that has a significant impact on customers. The cost of getting this wrong could be steep, both in fines and in lost trust. For startups building AI-powered financial tools, the takeaway is to bake transparency into your product from day one. Don't wait for regulators to force you into it. Make it easy for users to understand why a decision was made, and give them a straightforward way to challenge it. That's not just good ethics, it's good business. The debate over AI in banking is only going to get louder. Durham's research is a shot across the bow, and it's probably not the last one. Whether we see actual bans or just tighter rules, one thing is clear: the era of unchecked algorithms in finance is coming to an end.