Sponsored

The Financial Conduct Authority has reopened its AI Input Zone and set a hard close of 19 June 2026 for responses. For UK banks, insurers, asset managers and payments firms that are not inside the regulator’s eight-firm AI Live Testing cohort, this short consultation window is, in practical terms, the last public on-ramp into the standard the FCA is now building.

The Input Zone is described by the FCA as “one component of our AI Lab”. Its purpose is to gather concrete examples of good and poor AI practice across regulated financial services and to feed a good-and-poor-practice publication the FCA expects to deliver “later this year”. That publication, once it lands, will sit alongside Consumer Duty, SYSC governance expectations and existing model risk guidance as the de facto reference point UK supervisors will use when they ask firms how their AI works in production.

The window to influence what counts as “good practice” closes on 19 June. There is no second consultation pencilled in before the publication is drafted.

What the FCA is actually asking for

The Input Zone is not a sandbox and not a beauty parade. The regulator is asking firms, vendors, academics and civil society to submit live examples against a tightly framed scope. In the FCA’s own words, it wants evidence on:

The deliberate plainness of that language matters. The FCA is not yet committing itself to a prescriptive rulebook on AI; it is collecting the raw material it will use to decide whether one is needed and, if so, where. Firms that engage now help shape that judgement. Firms that do not, inherit whatever consensus the engaged firms and their advisers manage to set.

Why this is distinct from AI Live Testing

The Input Zone and AI Live Testing are often discussed in the same breath and the distinction is worth pinning down precisely.

AI Live Testing is the FCA’s closed cohort programme. The second cohort, announced on 21 April 2026, contains eight firms: Barclays, Experian, Lloyds Banking Group (via Scottish Widows), UBS, Aereve, Coadjute, GoCardless and Palindrome. Their AI systems are being put through structured testing under real workloads, with a partner assurance provider. The cohort’s testing concludes by year-end and an evaluation report is scheduled for Q1 2027.

That cohort is closed. Other firms cannot apply into it on a 19 June timetable. What the rest of the market can do — and what the FCA is explicitly inviting — is submit through the Input Zone. Both feeds run into the same compliance trajectory:

The good-and-poor-practice publication will draw from both. Firms inside the cohort have a structural advantage in shaping it; the Input Zone is the only mechanism by which outsiders can match that influence.

The risk of staying silent

UK financial regulation on AI is moving from advisory tone to delivery tone. The Mills review into advanced AI in retail services is feeding into recommendations due in the summer. The PRA’s 2026/27 business plan signals continued monitoring of AI use across PRA-regulated firms. Consumer Duty is already being applied to AI-driven outcomes in pricing, claims and credit decisions.

In that environment, the good-and-poor-practice publication will not read as guidance in the soft sense. It will read as the benchmark against which supervisors test board reports, internal audit findings, complaints data and incident logs. Anything that diverges from it will need a clearly evidenced reason.

For firms outside the cohort, three risks follow from not engaging with the Input Zone:

  1. Standards calibrated for other firms. If only large-bank governance models, only insurer-style model risk frameworks or only large-vendor assurance practices reach the FCA, the published expectations will tilt to those shapes. Smaller asset managers, building societies and specialist lenders that deploy AI differently may find themselves judged against a template that does not match how they actually operate.
  2. No second bite before publication. The Input Zone is open until 19 June and there is no signalled re-opening before the FCA writes up its findings. Comment-on-draft routes will exist, but the framing will already be set.
  3. Weaker bargaining position on later rules. If the good-and-poor-practice document does prompt formal rule-making — for example on AI incident reporting, override design or third-party model assurance — the consultation responses that carry weight will be those grounded in submissions the FCA has already seen.

What a useful submission looks like

The FCA is explicit that it wants examples, not position papers. The most useful submissions will:

Firms with mature model risk functions will find much of this material already exists in internal documentation. The harder editorial work is deciding what to share, what to anonymise, and how to write in language that is useful to a regulator that will publish a synthesis, not the underlying submissions.

The narrow window

The trajectory the FCA has set out — Live Testing through 2026, evaluation in Q1 2027, good-and-poor-practice publication later in 2026 — does not leave much space for firms to argue their corner after the fact. The Input Zone is the public-facing edge of that trajectory and it closes on 19 June 2026.

For UK financial services AI compliance leads, the next three weeks are the cheapest opportunity they will have all year to put their own evidence in front of the regulator before the standard is written.

Primary sources

Finance & Markets Correspondent
Covers: Finance, capital markets, technology investing

David Whitmore covers the intersection of capital and code — the funding rounds, market structures and policy moves that shape how money flows through the technology economy.