The Financial Conduct Authority cited 86 percent AI adoption from a survey due out the following week. Two of its 'immediate' AI actions were already running before the review meant to prompt them.
Estimated reading time: 6 minutes
.jpg){kind=link}
In short. On 19 October 2026 the FCA's chief data, information and intelligence officer, Jessica Rusu, told Bloomberg's UK Policy Series that "our joint AI survey with the Bank of England, published next week, shows 86% of firms now use AI" [1]. The survey itself had not been published at the time of the speech, and still was not the next day. Its predecessor, the third joint Bank of England/FCA AI survey, published 21 November 2024, split its headline finding in two: 75% of firms already using AI and a further 10% planning to within three years [2], a structure the new single 86% figure does not obviously map onto. Separately, Rusu said the FCA is "embedding 6 of the 7 recommendations into live work immediately", pointing to "innovation services such as the AI Lab" as evidence [1]. The AI Lab was first published 13 October 2024, two years before the Mills Review that supposedly prompted its "scale up" [3], and the Handbook API she also cited launched 6 August 2026, more than two months before this speech [4]. Two documented mistakes: four stars.
Picked from this week's regulator feed: an FCA speech published yesterday, verified live on fca.org.uk with no withdrawal notice, scoring on three criteria, sector weight, high search salience and a first-of-kind call to AI providers with no enforcement mechanism attached. This is a Tier 2 structured teardown: the speech's own words tested against the documents it cites, and against when those documents actually date from.
A regulator addressing "humans" about an AI future it says has already arrived
Rusu opened by telling her audience it was "good to be here today with you, humans", before setting out how the FCA sees AI reshaping the sector it regulates: global AI spend "expected to reach over $2.6 trillion this year", ChatGPT past a billion active users, and firms "moving from experimentation to deployment" [1]. The speech ran through five areas: the FCA's response to the Mills Review, new "good and poor practice" guidance published the same day, the shift to "agentic finance", the FCA's own use of AI as a "smarter regulator", and a call on AI providers to add disclaimers to the financial information their tools generate [1]. None of it amounts to a new rule. Tristram Lawton, a financial services regulation partner at Simmons & Simmons, told Law News the absence of new rules "was expected": "The FCA is not introducing any new rules for firms using AI to deliver financial services. This is not a surprise" [5]. What the speech asserts instead is a set of facts and a pace of action, and both are where the trouble sits.
The claims, tested
| The speech's claim | What we found | Verdict |
|---|---|---|
| "Our joint AI survey with the Bank of England, published next week, shows 86% of firms now use AI" [1] | As of the speech, and still as of this piece, no 2026 edition of the survey appears on either the FCA's or the Bank of England's websites. The 2024 edition measured two separate figures, 75% using and 10% planning, up from 58% and 14% in 2022 [2], a structure the new single 86% headline does not match | A precise statistic, delivered to a live industry audience, with no published methodology, sample size or comparable breakdown available to check it against |
| "We are embedding 6 of the 7 recommendations into live work immediately. Our innovation services such as the AI Lab give us direct insight into what firms are building" [1] | The Mills Review was published 6 July 2026 [6]. The AI Lab it credits was first published 13 October 2024 [3]; the Handbook API she names a sentence later launched 6 August 2026 [4]. Both pre-date the Review | Work already running before the Review is credited as evidence of recommendations "embedded... immediately" after it |
A number before its evidence
The 86% figure is the kind of statistic built for a headline, and it got one: inside a day, trade press was running "FCA reveals 86% of financial firms now use AI" off the speech alone. The FCA has run this survey three times. The first, in 2022, found 58% of firms using AI and 14% planning to. The second, published 21 November 2024, found 75% using and a further 10% planning, a jump the Bank's own report called out explicitly as "higher than the figures in the 2022" survey, alongside detail on sector split (insurers highest at 95%, financial market infrastructure firms lowest at 57%), third-party exposure, and a sample of 118 respondents across six sectors [2]. That level of detail is what gives a number like "75%" something to stand on. Rusu's "86%" was delivered, by her own account, a week ahead of the document that would let anyone check what it measures, how many firms answered, or whether "now use AI" covers the same ground as "already using AI" did in 2024. A regulator that spent its 2024 report carefully distinguishing current use from future plans gave its 2026 audience one compressed number and asked them to take it on trust until the following week.
"Immediate" work that predates the review crediting it
The Mills Review's own page lists its seven recommendations plainly: secure the regulatory perimeter, strengthen coordination, monitor the shift to autonomous models, scale up the AI Lab, enable agentic finance, build an agentic supervisory model, and develop a public-interest AI financial capability service [6]. Rusu's speech says six of the seven are going "into live work immediately", and names the AI Lab and the Handbook API as proof of momentum [1]. Neither is new work. The AI Lab's own page records it as first published 13 October 2024, with a Supercharged Sandbox cohort that had already completed its programme and an AI Live Testing cohort running since April 2026, all before the Review that recommended "scaling it up" existed [3]. The Handbook API, cited in the same breath as part of "dynamic data collection" supervision, launched 6 August 2026, over two months before this speech and a month after the Review was published, built by the FCA's Alex Smith and team as a general compliance tool with no stated connection to Mills at launch [4]. Crediting an "immediate" response to a review with initiatives that were already running, one of them for two years, is not evidence of speed. It is evidence that the review found work the FCA could point to, which is a different thing.
Credit where due
Not everything in the speech blurs its own timeline. On the recommendation it is not following as written, a new public-interest financial capability tool, Rusu said so directly: "rather than create a new public-interest financial capability tool, as the Mills Review suggested, we will build on existing work" [1], a disclosed substitution rather than a buried one. The "good and poor practice" document published alongside the speech is substantive, not padding: it sets out real examples, a firm building a live list of vulnerability signals for its chatbot, another running an independent ethics panel before deployment, and genuine poor practice, firms excluding vulnerable consumers by default or retrofitting governance after a system has already gone live [7]. And the AML supervision work the speech describes, building AI tools to oversee 60,000 firms entering the FCA's remit in 2028, rests on a reform that is genuinely confirmed and under way, not invented for the speech [1].
Verdict
Four stars, from two documented mistakes. Neither is a claim that the FCA is wrong to focus on AI, or that its good and poor practice guidance is poorly made. The first mistake is a precise public statistic delivered a week ahead of the data that would let anyone check it, breaking with the FCA's own more careful precedent of publishing adoption figures alongside sample size and methodology. The second is crediting "immediate" action on a review to initiatives that were running before the review existed, one of them by two years. A regulator calling for firms to maintain "clear accountability" and resist "overstating capabilities" [1] in their own AI claims should hold its own announcements to the same test.
The star score counts two documented mistakes: the "86% of firms now use AI" figure, attributed to a joint Bank of England survey that remained unpublished at the time of the speech and the day after, with no methodology or sample available to check it against, and materially different from the two-part measure (75% using, 10% planning) its 2024 predecessor used; and the claim of "embedding 6 of the 7 recommendations into live work immediately", evidenced by the AI Lab (first published 13 October 2024) and the Handbook API (launched 6 August 2026), both of which pre-date the 6 July 2026 Mills Review credited with prompting them. Two falls in the 1 to 3 band: four stars; the bands are on the ratings page. Checked directly against the FCA speech transcript, the Mills Review page, the AI Lab page, the Handbook API blog post and the Bank of England's 2024 AI survey report, fetched 20 October 2026. The document is tracked in the Daily Register; a revision, or the survey's eventual publication, will get a diff entry.
Sources
- Building the future of finance in the age of AI, Jessica Rusu, FCA, speech delivered 19 October 2026
- Artificial intelligence in UK financial services - 2024, Bank of England and FCA, published 21 November 2024
- AI Lab, FCA, first published 13 October 2024, last updated 3 July 2026
- Making compliance simpler: opening up the FCA Handbook through our new API, Alex Smith, FCA, published 6 August 2026
- FCA sets no new AI rules but tells senior managers they will be held to account, Law News, 19 October 2026
- AI and the future of retail financial services (The Mills Review), FCA, published 6 July 2026
- AI in financial services: good practice and areas for improvement, FCA, published 19 October 2026