FCA AI approach: what the AI Lab offers firms deploying AI
In short: the FCA AI approach pairs no new AI rules with supervised engagement. The Consumer Duty, the SM&CR and existing governance expectations govern AI in UK financial services, while the AI Lab offers five engagement routes. The AI Input Zone closes on 19 June 2026, feeding a good and poor practice publication later in 2026.
There is no AI rulebook in UK financial services, because the FCA AI approach applies the rules that already exist. What the regulator offers instead is engagement: testing in live markets under supervisory oversight, computing power for early-stage development, and an open survey asking where clarity would help. That survey, the AI Input Zone, closes on 19 June 2026.
The Financial Conduct Authority (FCA) restated the position in a blog by Alex Smith, its Head of Cross-cutting Policy and Strategy, on 2 June 2026. The FCA committed in its response to the Prime Minister’s letter on growth-supporting regulation to avoid additional regulation for AI by relying on existing frameworks. The blog confirms what comes next: a good and poor practice publication later in 2026, and the independent Mills Review into the long-term impact of AI on retail financial services, due in the summer.
What the FCA AI approach relies on instead of new rules
The FCA applies three existing frameworks to artificial intelligence: the Consumer Duty, the Senior Managers and Certification Regime (SM&CR) and its expectations on governance and controls. The Consumer Duty, made through Policy Statement PS22/9 as PRIN 2A of the FCA Handbook and in force for open products since 31 July 2023, requires firms to act to deliver good outcomes for retail customers. Its four outcomes do most of the work on AI: consumer understanding governs AI-generated communications, price and value governs AI pricing models, and consumer support governs automated triage of customers in vulnerable circumstances.
The SM&CR, built on the Financial Services and Markets Act 2000 (FSMA 2000), attaches individual accountability for AI systems to named senior managers. Its scope matters for payments: the regime binds FSMA-authorised persons, and payment institutions and e-money institutions sit outside it, as set out in SM&CR reform and payment firms. For those firms the FCA overlay on AI is the Consumer Duty and governance expectations.
None of this displaces data protection law. AI systems processing personal data must satisfy the UK GDPR, including the automated decision-making conditions inserted by the Data (Use and Access) Act 2025 and in force since 5 February 2026. The FCA’s supervisory expectations sit on top of those obligations, not in place of them. The four-regulator picture, including the parallel work of the ICO, CMA and Ofcom, is set out in AI consumer protection UK.
The FCA AI approach in practice: five AI Lab routes
The FCA AI Lab is the regulator’s engagement programme for artificial intelligence, with five components. The Supercharged Sandbox, run in partnership with NVIDIA, gives firms high-performance computing, enriched datasets and advanced AI tools for early-stage development; the first cohort completed the programme and presented at a showcase on 28 and 29 January 2026, and applications for the second cohort are under consideration. AI Live Testing gives firms further along in AI development a way to test systems in real-world conditions with regulatory oversight, supported by Advai, the FCA’s technical partner in automated AI assurance. The second cohort, announced on 21 April 2026, comprises eight firms, including major retail and investment banks, a credit reference agency and payments firms; their use cases include agentic payments, anti-money laundering detection, Know Your Customer checks and AI-enabled targeted support for investments. Testing concludes by the end of 2026, with an evaluation report due in Q1 2027.
The AI Spotlight showcases how firms experiment with AI in financial services; applications remain open, and Spotlight firms gain accelerated entry to the Monetary Authority of Singapore’s pathfin.ai initiative under the FCA’s partnership with MAS. The AI Sprint, held in January 2025, brought industry, academics, regulators and consumer representatives together to inform the FCA’s regulatory approach. The AI Input Zone is an open survey on good and poor practice in AI use, with responses due by 19 June 2026.
| AI Lab component | What the FCA provides | Where it stands at 10 June 2026 |
|---|---|---|
| Supercharged Sandbox | High-performance computing, enriched datasets and advanced AI tools, in partnership with NVIDIA | First cohort completed; second-cohort applications under consideration |
| AI Live Testing | Testing of AI systems in live markets with regulatory oversight, supported by Advai | Second cohort of eight firms testing from late April 2026; evaluation report due Q1 2027 |
| AI Spotlight | A showcase of AI use cases, with accelerated entry to MAS’s pathfin.ai initiative | Applications open |
| AI Sprint | A convening event informing the FCA’s regulatory approach | Summary published; may be repeated |
| AI Input Zone | An open survey feeding a good and poor practice publication later in 2026 | Closes 19 June 2026 |
What participation offers payment and e-money firms
Participation in the AI Lab is voluntary and confers no regulatory approval. It gives firms early sight of supervisory expectations before the FCA writes them down. The FCA’s own framing of AI Live Testing is clarity on expectations, confidence in deployment and a chance to shape best practice. The FCA asks applicants for a developed proof of concept, deployment plans, and a willingness to share their testing approach and findings. The second cohort’s use cases show where the FCA expects AI to concentrate in payments: agentic payments, financial crime screening and customer onboarding all engage the Consumer Duty on outcomes.
For a payment service provider, e-money institution or fintech weighing engagement, the dated points under the FCA AI approach are concrete. AI Input Zone responses are due by 19 June 2026, and the FCA anonymises examples before publication. The good and poor practice publication later in 2026 will be the closest thing UK financial services has to AI guidance under the current no-new-rules posture. AI Spotlight applications remain open. The wider supervisory regime for payment firms is summarised on the payments regulation page, and the data protection workstream that sits underneath any AI deployment, including AI governance and impact assessments, is set out in the AI and data governance advice guide.
Viewpoint
The AI Lab is engagement, not approval, and the distinction matters. A firm that completes AI Live Testing still has to satisfy the Consumer Duty on outcomes and the UK GDPR on processing; the FCA AI approach has been consistent on that point since the regulator’s response to the Prime Minister’s letter. In our experience advising payment and e-money firms on AI-enabled products, the difficult questions sit below the FCA overlay: the consent architecture for agentic payments under regulation 67 of the Payment Services Regulations 2017, examined in agentic AI and payment authorisation, and the automated decision-making conditions under the UK GDPR. I read the second cohort’s use-case list, with agentic payments and financial crime screening at its centre, as the clearest signal yet of where FCA supervision of AI will concentrate. Three dated outputs will frame the next twelve months: the good and poor practice publication later in 2026, the Mills Review in the summer, and the AI Live Testing evaluation report in Q1 2027.
Frequently asked questions
What is the FCA AI approach to new AI rules?
The FCA AI approach relies on existing frameworks rather than new regulations. The FCA confirmed on 2 June 2026 that it will not introduce new rules for AI; the Consumer Duty, the SM&CR and its expectations on governance and controls govern how firms use the technology in regulated financial services.
What is the FCA AI Lab?
The FCA AI Lab is the engagement arm of the FCA AI approach. It has five components: the Supercharged Sandbox, AI Live Testing, the AI Spotlight, the AI Sprint and the AI Input Zone. Participation is voluntary and does not confer regulatory approval.
When does the AI Input Zone close?
The AI Input Zone closes on 19 June 2026. It asks three questions: what has enabled safe and responsible AI use cases, what has stopped them, and what themes the FCA’s good and poor practice publication needs to address. The FCA anonymises responses before the publication later in 2026.
Does AI Live Testing give firms regulatory approval?
No. AI Live Testing provides a setting to test AI systems in real-world conditions with regulatory oversight, but it confers no approval and does not relieve a firm of its obligations under the Consumer Duty, the SM&CR or the UK GDPR. The second cohort’s evaluation report is due in Q1 2027.
For advice on deploying AI in a regulated payments or e-money business, contact Rob Bratby at Bratby Law.
