This case study examines how AI tools respond to regulatory questions relevant to Compliance teams at Mutual Funds and UCITS firms operating in the United Kingdom. It covers one regulation — the Consumer Duty (PS22/9 and PRIN 2A) issued by the Financial Conduct Authority — across one aggregated question where AI assistants produced materially incorrect answers. Across the findings reviewed, AI tools consistently misrepresented the FCA's stated expectations, introducing obligations and framings that are not present in the rule text.
Compliance teams relying on these AI responses without independent verification would be working from a distorted picture of what the regulator actually requires.
Compliance teams at Mutual Funds and UCITS firms routinely encounter Consumer Duty questions in the ordinary course of their work. Fair value assessments sit at the centre of many of those enquiries: teams drafting product governance frameworks, preparing annual value assessments, or supporting distribution arrangements need to understand precisely what the FCA expects firms to do — and equally, what the FCA does not require.
When a Compliance officer or a business-line stakeholder turns to an AI tool to get a quick steer on these questions, the output shapes internal policies, training materials, committee papers, and distribution-review processes before anyone has checked it against the actual regulatory text.
The downstream uses are significant. A fair-value methodology built on incorrect assumptions about FCA expectations can embed itself in a firm's product oversight and governance framework, in pre-contractual disclosures reviewed by distribution partners, and in the supporting documentation provided to the board or to an independent Non-Executive Director with Consumer Duty oversight responsibility. Supplier due-diligence questionnaires sent to third-party administrators or distributors may also carry forward the same error if they are drafted using AI-generated descriptions of the firm's obligations.
The firm, not the individual employee, bears the regulatory and financial consequences when these errors surface. The FCA has broad powers under the Consumer Duty to require remediation, impose fines, and publish censures. Where a firm has embedded an incorrect understanding of its fair-value obligations into live products or distribution arrangements, the cost of correction — scope review, retrospective analysis, potential client redress, and engagement with the regulator — can be substantial. The Compliance team's credibility with the board, with distribution partners, and with the regulator is also at stake.
The findings in this case study converge on a single, consistent pattern: AI tools upgrade or invert the FCA's stated position on what firms must do. Where the regulator has deliberately set a ceiling — explicitly stating it does not expect a particular level of analysis — AI assistants tend to raise that ceiling, reframing the FCA's restrained expectation as an encouragement or a conditional obligation. The effect is that a Compliance team reading the AI's answer would believe it needs to do more, or to structure its work differently, than the actual rule requires.
This is not a failure to find the right source; it is a failure to read the source accurately.
All findings in this review relate to a single regulation — the Consumer Duty as set out in PS22/9 and PRIN 2A — and cluster specifically around fair-value methodology. The FCA spent considerable effort in its final policy statement distinguishing between what firms must do and what it is not prescribing. AI tools tested on this topic consistently collapsed that distinction, treating the regulator's permissive or limiting language as though it were a floor rather than a ceiling. That is a structurally important error in a rule designed, in part, to give firms flexibility in how they demonstrate value.
The systemic risk for a Mutual Funds or UCITS firm is that the same incorrect AI answer feeds multiple downstream work-products. A single AI-assisted description of fair-value methodology, if unverified, can propagate through product governance documentation, annual assessment templates, board reports, and distribution agreements in a single review cycle. Each of those documents then requires correction if the error is discovered — and the further the error has travelled, the more expensive and disruptive the remediation becomes.
Where the same AI tool is consulted by more than one team member independently, the error may also be reinforced rather than caught, because multiple people receive the same incorrect confirmation.
1 finding in this case study. Click any to see its full evidence card.
The default position for any Compliance team at a Mutual Funds or UCITS firm should be that AI tools are a starting point for orientation, not a primary source for determining what a regulation actually requires. This is especially true for Consumer Duty obligations, where the FCA has taken care to define both what it expects and what it is not prescribing — and where AI assistants have consistently failed to preserve that distinction accurately.
Before any AI-generated description of regulatory obligations enters a firm work-product, it should be verified against the source regulatory text or authoritative FCA guidance by someone with the relevant competence to evaluate it.
At a firm level, practical safeguards include a written policy that identifies AI tools as unreliable primary sources for rule interpretation in Consumer Duty and related regulatory areas, with a named sign-off requirement before AI output influences any regulatory-facing document, board paper, or client-facing disclosure. Where AI output has been used to draft or shape a work-product, that should be documented — both to create an audit trail and to support quality-review processes.
Teams should also distinguish clearly between content that has been AI-drafted and subsequently verified, and content that has merely been AI-summarised without independent checking; those two categories carry different levels of regulatory risk and should be treated differently in governance processes.
There are areas where AI tools can be used more safely within a Compliance workflow. Drafting non-regulatory internal communications, generating first-draft questions for further legal or regulatory research, or summarising long documents that the team will then read and verify are all lower-risk uses. The key variable is whether the firm is relying on the AI to determine what the rule says. Where the answer to that question is yes, independent verification is not optional.
RegLeg publishes its hallucination research openly so that Compliance teams can use it as a free reference check before relying on AI answers in regulated areas. For a Compliance team at a Mutual Funds or UCITS firm, the Consumer Duty findings documented here are a direct signal: if your team has used AI tools to inform fair-value methodology documentation, product governance frameworks, or distribution-related assessments, checking those outputs against the FCA's actual text is a practical first step.
The published research identifies the specific questions where AI tools have been shown to go wrong, which means teams can prioritise their verification effort rather than reviewing everything from scratch.
Beyond the published research, RegLeg offers bespoke regulatory deep-dives for Mutual Funds and UCITS firms that map AI-supported workflows against the hallucination patterns observed across the Consumer Duty and related rule areas. This is particularly useful where a firm's Compliance team has already embedded AI tools into regular processes — product oversight reviews, annual value assessments, distribution-partner due diligence — and wants to understand where the highest exposure sits before a regulator asks the same question.
The output is practical: a prioritised view of which workflows carry material risk and which are lower-exposure, so the team can concentrate its verification and governance resources where they matter most.
RegLeg also offers confidential review of a firm's existing AI-use policy against our failure-mode catalogue, with prioritised remediation guidance tailored to the Consumer Duty obligations most relevant to a Mutual Funds or UCITS Compliance function. For teams that need to build internal capability, we can support the development of training material and CPD-aligned content explaining where and why AI tools fail on regulatory interpretation — content that Compliance teams can use directly with business-line colleagues, product teams, and senior management.