FCA Urges Financial Firms to Strengthen AI Safety and Oversight

The UK Financial Conduct Authority (FCA) has called on financial firms to use artificial intelligence (AI) safely, with clear responsibility, strong oversight, and plans for when systems fail.

Jessica Rusu, the FCA’s Chief Data, Information and Intelligence Officer, most recently explained in a speech how AI is changing financial services and the regulator’s own work.

The FCA also published examples of good and poor AI practices based on its work with firms. The review focuses on three areas: protecting customers, making sure someone is responsible for AI decisions, and keeping services running when technology fails.

What does the FCA expect from financial firms?

1. Put customers first

AI should improve customer service, not simply automate existing processes. The FCA highlighted good practices such as allowing customers to speak to a human adviser when needed and using AI to identify customers who may require extra support. It also warned against deploying tools without a clear purpose, overstating what AI can do, or failing to properly test automated customer journeys.

2. Make responsibility clear

Firms should determine who is responsible for their AI systems and the decisions they support. The FCA expects appropriate controls, human oversight, clear escalation processes, and senior-level accountability. It warned that some firms introduce governance measures only after AI tools have already been deployed. The regulator says firms should address risks proactively before systems go live and keep monitoring them as they change.

3. Prepare for failures

AI systems can make mistakes as they are used more widely or as the underlying models change. The FCA says firms should test their systems under pressure, maintain backup options, and understand their dependence on technology suppliers. The regulator also stressed the importance of reporting major incidents, AI failures or near-misses.

The FCA plans to use AI in its own supervision

The regulator is also applying AI to its work. It is developing tools to help oversee around 60,000 anti-money laundering (AML) firms expected to come under its remit in 2028. It is also building AI tools to help supervise around 9,000 smaller firms that report transactions in wholesale markets.

According to the FCA, these tools could help staff review information faster, spot early warning signs, and focus on cases that need close monitoring.

The message is clear that AI can help financial firms and regulators work more effectively, but it must be supported by clear responsibility, customer protection, proper testing, and reliable backup plans.

Implications for Compliance Teams

• Proactively review AI tools before launch: Confirm that each system has a clear purpose, appropriate testing, and controls suited to its risks.
• Keep human check: Set clear rules for when staff must review, challenge, or take over an AI-supported decision.
• Monitor systems after deployment: Check for changes in performance, unexpected results, and risks as models or data change.
• Prepare for disruptions: Maintain backup procedures and understand the risks created by dependence on external AI suppliers.
• Keep records and escalate problems: Document decisions, assign responsibility, and follow reporting requirements when serious failures occur.

The practical lesson for AML teams is that AI can support screening, monitoring, and investigations, but it is the responsibility of firms to check how these tools are used and how risks are managed.