UK financial regulator demands firms report AI failures and near misses
The FCA chief data officer says enterprises must disclose significant AI errors and close calls to protect customers and market integrity.
Key points
- The UK Financial Conduct Authority wants mandatory reporting for significant AI failures and near-misses.
- Jessica Rusu states that agentic AI systems require continuous assurance rather than one-off safety checks.
- Regulators are frustrated by rapid AI rollouts that lack meaningful consequences for misbehaviour.
Customers and clients of UK financial firms face new transparency rules regarding artificial intelligence. The Financial Conduct Authority has stated that enterprises must report significant AI failures and near-misses. This directive targets the growing use of agentic and generative AI in financial services. The regulator aims to address concerns about system behaviour that may change under stress or at scale.
Who is affected by this new reporting requirement?
The requirement applies to enterprises operating within the UK financial sector. According to CSO Online, Jessica Rusu, the FCA Chief Data Officer, outlined these expectations during a speech to a Bloomberg UK Policy Series audience. The rules extend existing major incident reporting obligations to include AI-specific issues. This means banks, insurers and other regulated firms must disclose when their AI systems cause serious harm or come close to doing so.
What counts as a significant failure or near-miss?
The FCA has not yet provided a precise definition for these terms. Questions seeking clarification were unanswered by publication time. However, Pieter Arntz from Malwarebytes suggests a significant failure involves serious consequences. In financial services, this could mean authorising fraudulent payments or exposing sensitive customer information. It might also include denying access to essential services or triggering transactions that threaten market integrity.
How do regulators view the current state of AI deployment?
Regulators express frustration over the rapid rollout of agentic and generative AI. Rusu noted that attempts to slow these deployments have often been futile. Reports of agent misbehaviour continue to emerge with alarming frequency. She emphasised that a system safe at deployment may behave differently later. Therefore, agentic systems require continuous, conditional assurance rather than static safety checks.
What framework is being developed for agentic finance?
The FCA plans to work with industry peers, fellow regulators and frontier labs. Their goal is to develop a trusted framework for agentic finance. This framework will include clear approaches to agent identity and delegated authority. It will also address accountability and execution standards. The aim is to create meaningful consequences for AI misbehaviour and improve operational resilience across the sector.
What affected people can do
Customers should monitor their financial accounts for unauthorised transactions or unexpected denials of service. If you notice irregularities that might stem from automated decision-making, contact your provider immediately. Be wary of follow-up communications claiming to resolve AI-related issues, as scammers often exploit such news. Verify any requests for personal data through official channels. Check whether your bank has updated its privacy policy regarding AI usage and data processing.
What to do and how to stay safe: FCA
- Review your recent bank statements for any transactions you did not initiate or recognise.
- Contact your financial institution directly if you experience sudden access issues or service denials.
- Enable multi-factor authentication on all financial accounts to add an extra layer of security.
- Monitor official communications from your provider for updates on AI system incidents or outages.
Step-by-step guide: Is It Safe to Use Password Managers? The Real Risks and Rewards
General security guidance from the DarkWeb Dispatch newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Does the FCA require reporting of all AI errors?
No, the FCA specifically requires reporting of significant AI failures and near-misses, not minor technical glitches.
Has the FCA defined what constitutes a significant failure?
The FCA has not yet provided a precise definition, though experts suggest it involves serious consequences like fraud or data exposure.
Who made these comments about AI reporting?
Jessica Rusu, the Chief Data Officer at the UK Financial Conduct Authority, made these comments during a Bloomberg speech.
