• The FCA is calling on AI providers to use clear, proportionate and consistent disclaimers; nothing found so far shows a binding disclaimer rule.
• The FCA says warnings may work best at the point where AI-generated financial information is delivered, not buried in terms of use.
• Consumer research cited in the FCA speech found that only 40% of people correctly recognise there is no recourse if something goes wrong after acting on a general-purpose AI tool's advice.
• The Mills Review flags reduced consumer agency, reduced understanding and unconscious manipulation as risks as AI agents begin to act for consumers.
• Compliance teams should map each AI feature to the activity it performs and place any warning at the point of delivery, while waiting for formal FCA detail.

The UK's financial watchdog wants AI providers to tell consumers plainly where machine-generated financial guidance stops being reliable. Finextra reports that the FCA is calling for disclaimers on services that give financial advice to consumers through agentic AI. The FCA's own wording, in a speech on AI in finance, is a call on providers rather than an instruction, and no binding disclaimer rule has been identified. The design question is who carries the risk when a consumer acts on a model's output.

A request to providers, not yet a rule

In the speech, the FCA says warnings may work best at the point where AI-generated financial information is delivered, and that it is calling on AI providers to help consumers understand the limits of that information through clear, proportionate and consistent disclaimers. That is the language of expectation. The material reviewed shows no rule or consultation that fixes the wording, placement or enforcement of such a disclaimer, so any firm treating it as settled compliance text is ahead of the evidence.

Why the regulatory perimeter is the open question

A voluntary disclaimer expectation leaves the harder question unanswered: which AI products are regulated activity and which are not. The sources reviewed do not draw that line for agentic tools, and this piece does not assume where it falls. The incentive is still structural. A provider that positions a product as general-purpose may expect lighter obligations than one that markets itself as giving advice, and a disclaimer is the cheapest signal of which side a product claims.

The consumer gap the disclaimer targets

The FCA speech cites consumer research finding that only 40% of people correctly recognise there is no recourse if something goes wrong after they act on advice from a general-purpose AI tool. That gap between assumed and actual protection is the mechanism a disclaimer targets. Whether a warning changes behaviour, rather than being ignored, is not established by the sources reviewed, so the request is best read as a hypothesis the regulator is testing.

Agentic delegation raises the stakes

The Mills Review notes that AI agents may increasingly construct, rebalance and execute investment portfolios for consumers, and flags reduced consumer agency, reduced understanding and unconscious manipulation as risks. When an agent acts rather than informs, a disclaimer shown at the start of a conversation is a weak control. The second-order effect for compliance teams is pressure to define which actions an agent may take without a fresh, informed confirmation.

What compliance teams should do now

The FCA speech describes the Consumer Duty as an outcomes-based framework and gives AI examples within it, which suggests regulators will judge outcomes rather than disclaimer wording alone. Practically, teams should map each AI feature to the activity it performs, document where consumers could mistake output for regulated advice, and place any warning at the point of delivery. They should then watch for a formal FCA consultation that turns the request into text.