The FCA used a 19 October speech to ask AI providers for “clear, proportionate and consistent disclaimers” on financial information and advice their models generate. The same speech published examples of good and poor AI practice at regulated firms and said major-incident reporting must reach AI failures and near-misses.
The disclaimer request
Jessica Rusu, the FCA’s chief data, information and intelligence officer, gave the speech at Bloomberg’s UK Policy Series. She said consumers are turning to general-purpose large language models such as ChatGPT, Gemini and Claude for financial information and advice, while regulated firms begin to integrate their services into AI platforms. That can widen access, she said, and it can also blur the boundary between regulated and unregulated activity.
She gave two figures. Four in five less-experienced investors aged 18 to 40 have used AI to help with investing, but only 40% of consumers correctly recognise that there is no recourse if something goes wrong after acting on advice from a general-purpose AI tool. The FCA says it will clarify the existing boundary instead of simply expanding its remit. Rusu said “warnings may be more effective at the point where AI-generated financial information is delivered”, and the FCA is therefore calling on AI providers to add disclaimers there.
Advertisement
The FCA is not taking up one recommendation from the Mills Review, which it commissioned and published in July to look at how AI could reshape retail financial services by 2030. The review suggested a new public-interest financial capability tool. The FCA will build on existing work, including an IOSCO TechSprint powered by its AI Lab to help investors spot scams, verify information and challenge unreliable AI outputs. Rusu said the FCA is embedding 6 of the 7 Mills recommendations into live work immediately.
Good and poor practice, drawn from firm testing
The FCA published its good and poor practice examples the same day. They come from testing it ran with firms this year and are organised around three themes: consumer centricity, accountability and governance, and operational resilience.
On consumers, the FCA cites a firm that built a live list of vulnerability signals its chatbot uses to flag customers who may need a human adviser. The poor examples it lists are tools built without a defined need, vulnerable consumers excluded by default, overstated capabilities and automated customer journeys deployed without enough testing.
On governance, Rusu said too many firms are still under-investing. The FCA expects clear accountability, risk-based controls, human oversight, escalation triggers and senior managers who are personally accountable for AI systems. Its good example is a firm that set up an independent ethics panel to examine proposed AI use cases before deployment, covering ownership, bias and controls. Its poor example is governance retrofitted after a system is already live. The FCA also names monitoring that is not adapted as systems change.
On resilience, the guidance asks firms to stress-test AI systems, keep fallback options and understand supplier concentration, including reliance on a single model. We covered the supplier side of that concern in The UK Just Extended Financial Regulation Beyond the Banks to the Cloud Providers Underneath Them. Rusu said the FCA will update the practice examples as the technology changes.
Agents and incident reporting
The speech then moves from models that help people decide to agents that act. Rusu described agents that can search across products, manage payments and investments, and transact directly with other agents, creating a new layer of autonomous technology between institutions and their customers. Meta’s Muse launch is her example of the shift already under way. Our analysis of what the coverage says about agent purchases on Stripe’s Link wallet looks at the payment side of the same shift.
Two commitments follow. First, Rusu said, “our rules already require firms to report major incidents, and that must extend to significant AI failures and near-misses”. She noted that the AI Minister has recognised near-miss reporting for agentic AI as a national security priority, and called it a regulatory priority too. Second, assurance has to change. She said “agentic systems therefore require continuous, conditional assurance” because a system that is safe at deployment may behave differently at scale, under stress or as the models beneath it change.
The FCA says it will work with industry, other regulators and frontier labs on a trusted framework for agentic finance, covering agent identity, delegated authority, accountability and execution. No date or draft was given. Our earlier reporting on the regulator’s sandbox work with AI developers is in Financial Regulators Are Recruiting AI Labs Directly Into Their Sandboxes, and the FCA Just Made Anthropic the Test Case.
The FCA’s own use of AI
Rusu spent the last part of the speech on the FCA as a user. It is building AI tools for oversight of the 60,000 anti-money laundering firms that come into its remit in 2028, and a “market first responder” agent that will supervise about 9,000 small wholesale markets transaction reporting firms so supervisors can focus on more complex work. The FCA launched a Handbook API in August that lets firms feed its machine-readable rules into their own AI tools, and it is working on machine-readable data collection APIs.
She also said a joint AI survey with the Bank of England, due next week, shows 86% of firms now use AI.
What it means for the finance leader
In our reading, three items in the speech carry dates or duties that a compliance team can act on now. The incident-reporting expectation applies under rules that already exist, so a firm with AI in a customer journey should check whether its incident taxonomy has a category for AI failures and near-misses. The practice examples show what the FCA saw in firm testing, and a firm that cannot name a senior manager for each AI system has a gap against the governance expectation.
The speech frames the disclaimer as a call on AI providers. Firms that integrate their services into AI platforms should still expect to be asked how consumers are told what protection applies at the point of advice, since the 40% recognition figure is the FCA’s stated reason for acting.
Source: FCA, “Building the future of finance in the age of AI” (speech by Jessica Rusu, 19 October 2026)
