When a central bank’s financial regulator opens its AI development program to Barclays, Experian, and Scottish Widows, the sandbox is no longer an experiment.
The UK Financial Conduct Authority has announced Cohort 2 of its AI Supercharged Sandbox, according to the FCA’s official announcement. Named participants include Aereve (compliance automation), Coadjute (property services), Barclays, Experian, and Scottish Widows. The participant list, sourced from the FCA’s own announcement, is confirmed contingent on URL resolution, if the FCA page lists these names, the confirmation is T1 authoritative.
Sandbox participants receive access to synthetic datasets and specialized compute resources provided by the FCA. The FCA’s announcement references specialized compute access; the specific involvement of NVIDIA in that compute provision is sourced from vendor partnership framing and should be treated as FCA-attributed only if the official announcement names NVIDIA directly. Until URL resolution confirms, the appropriate framing is: “specialized compute resources, according to the FCA.” Professional Adviser also covered the announcement.
Three things make Cohort 2 noteworthy beyond the headline names.
First, the participant mix signals breadth. Aereve and Coadjute are focused fintechs testing specific use cases, compliance automation and property services respectively. Barclays, Experian, and Scottish Widows are established institutions whose participation means the sandbox is handling the AI compliance questions that matter to mainstream financial services, not just to startups.
Second, the FCA’s model is structurally different from the EU’s approach. Brussels is building rules first and requiring industry to comply. The FCA is co-developing governance with industry participants, using live institutions and real use cases, and deriving regulatory lessons from what participants actually build. Neither approach is inherently better, but they produce different regulatory outputs at different speeds.
Third, this is the second cohort. The program existed before. It didn’t collapse. Cohort 2 is the FCA signaling that the first cohort produced results worth building on.
For compliance professionals in UK financial services, the FCA sandbox creates a meaningful reference point. The use cases being tested, personalization, compliance automation, property services, are adjacent to deployments already underway at institutions not in the sandbox. What emerges from the Cohort 2 evaluations will likely inform the FCA’s forward guidance on AI in financial services. Tracking what participants report, when they’re permitted to report it, is worth building into your regulatory monitoring calendar.
What to watch: FCA publications on Cohort 2 progress and findings. The sandbox’s value to the broader market is in what the regulator publishes about what it learns. Prior cohort outcomes, if published, provide a baseline for evaluating what Cohort 2 changes in the FCA’s thinking.
TJS synthesis: The FCA’s sandbox isn’t a soft approach to AI governance, it’s a different theory of governance. Rather than setting rules and enforcing compliance, the FCA is using structured experimentation with named institutions to build the evidentiary base for future regulation. Cohort 2’s roster suggests the model is working well enough for the UK’s largest financial institutions to want in. That’s a different signal than a regulator announcing new rules.