UK Lawmakers Urge Financial Services to Treat AI As a Systemic Risk

A UK parliamentary committee has warned financial services regulators that the current “wait and see” approach toward AI could expose consumers—and the broader financial ecosystem—to systemic risk.

The warning comes as financial services organizations, often in partnership with third-party fintechs, are operationalizing AI across an increasingly wide range of use cases, including some that have been deemed inherently high risk. Moreover, AI is no longer limited to assistive use cases, with enterprises now turning to agentic systems for end-to-end automation with minimal human oversight.

Specifically, the UK committee highlighted AI’s adoption in high-impact use cases like creditworthiness and credit scoring—areas highly susceptible to biased outcomes, not to mention myriad privacy and security risks. Many of the use cases discussed mirror those of Annex III of the EU AI Act, which applies to all organizations that do business in the EU, regardless of where they are located. The fear is that, when AI is used to inform or even automate these high-impact decisions, it could in turn increase the risk of model failure, opaque and unfair decisioning, and open up new fraud vectors.

Whether—or how—UK regulators translate this warning into formal testing regimes like auditability standards, governance controls, and scenario design remains to be seen. Nonetheless, the committee recommends that the Financial Conduct Authority (FCA) and the Bank of England start running so-called “AI stress tests” to prepare for risks that could be triggered by the deployment of increasingly automated systems.

While regulators the world over are struggling to keep pace with the rapid adoption of AI in financial services (the EU AI Act is still the only law of its kind), it is increasingly clear that financial services and fintech organizations must be proactive. Risk management, in the context of AI and automation, must shift from policies to proof. In other words, enterprises must be able to provide actual evidence of their monitoring, control, and explainability efforts, lest they be held accountable for failures.

What this ultimately boils down to is keeping humans in the loop by carefully monitoring model drift, applying kill switches, and ensuring that people are always able to override automated processes. Enterprises must also continuously evaluate vendor risk by mapping their critical dependencies and building in contingency plans. Moreover, AI teams and fintechs specializing in AI-powered solutions for financial services companies should treat explainability as a core product requirement. After all, you can’t hold a machine accountable when something goes wrong, whether it’s biased credit scoring, a failed fraud detection engine, or anything else.