TL;DR:

  • The Financial Conduct Authority has selected eight firms — Aereve, Coadjute, Barclays, Experian, Go-Cardless, Lloyds Banking Group (Scottish Widows), UBS and Palindrome — for the second cohort of its AI Live Testing initiative, running testing now through end-2026.
  • Use cases span AI-enabled targeted investment support, consumer credit score insights, agentic payments, anti-money-laundering detection and Know Your Customer — a deliberately broad mix across customer-facing and B2B deployments.
  • A Good and Poor Practice report is scheduled for publication later in 2026, with a full evaluation of the cohort due Q1 2027. Sandbox and Innovation Pathways applications were up 49% on the prior year.

Jessica Rusu, the FCA’s chief data, information and intelligence officer, announced the cohort at UK FinTech Week, positioning Live Testing as the regulator’s preferred mechanism for observing production AI inside regulated firms before formal guidance lands. The scale signal matters: eight firms on top of the seven already in cohort one make this the largest sandboxed AI-in-financial-services study by any Western regulator to date.

Breadth over depth in the case mix

The second cohort deliberately stretches across use cases rather than doubling down on any single one. Lloyds’ Scottish Widows arm is testing AI-enabled targeted support — the newly-created regulated category sitting between guidance and full financial advice — with the aim of addressing the UK’s advice gap. Coadjute will stress-test an AI-native anti-money-laundering platform aimed at the property market, timed to the forthcoming transfer of AML compliance from the SRA to the FCA. Barclays, UBS and Experian are using the cohort to surface agentic-payment and credit-scoring patterns the regulator has not yet ruled on. The diversity matters because it lets the FCA’s Good and Poor Practice report later this year speak to multiple AI workflows rather than a single narrow pattern.

What firms outside the cohort should take from this

The FCA has been consistent that AI Live Testing does not confer safe-harbour status — firms outside the cohort cannot cite cohort membership as a defence, but they can and should read the forthcoming practice report as supervisory guidance. Crucially, the FCA said in its Innovation Insights report that it intends to “avoid additional regulations for AI by relying on existing frameworks”. That positions Live Testing as the main instrument through which new AI-specific expectations will be communicated — via examples, not rules.

Looking forward

The 49% year-on-year increase in Sandbox and Innovation Pathways applications suggests AI Live Testing has created a procedural backlog the FCA is unlikely to clear without expanding capacity. UK firms contemplating customer-facing AI deployments in 2026 should plan around that queue: either apply for cohort three (expected late-year) or design deployments that can cite cohort-one findings when they are published. The separate Sheldon Mills review into advanced AI’s implications for consumers, retail markets and regulators — launched in January — will shape the forward posture alongside these practical outputs.

Share this article