Community Trust ScoreVerified
Anthropic is joining the FCA’s Supercharged Sandbox — and it’s not a small commitment. The AI company will hand participating firms access to two of its platforms, Claude Code and Claude Cowork, to speed up development across the second cohort of companies now working inside the regulator’s controlled testing environment.
The second cohort is 21 organizations strong. Names include Scottish Widows, Money Advice Trust, TrueLayer, calQrisk, Merx Digital Solutions Ltd (SmartDrops), Aegis Trace, Sardine AI Corp, Zquas, Trustie Labs, and Ubyx — the last of which is working with Amazon. It’s a wide spread of firms, from established financial names to newer entrants, all trying to push AI projects forward without stepping outside regulatory guardrails. The FCA received 199 applications for this second round, up from 132 in the first phase. That’s a 51% jump. Hard to ignore.
Not just Anthropic, either.
The Sandbox already had NayaOne and NVIDIA on board before Anthropic came in. NayaOne provides the foundational infrastructure — essentially the digital plumbing the whole thing runs on. NVIDIA brings computing power and its AI Enterprise software, which matters a lot when firms are trying to run serious model training or inference workloads. Anthropic’s addition layers in direct access to frontier AI tools, which is a different kind of resource than raw compute.
What Firms Are Actually Building
The use cases aren’t vague. Cohort participants are working on agent-led payments, fraud detection improvements, AI governance frameworks, and expanding access to financial services. Those are four pretty distinct problem areas, which probably explains why the applicant pool grew so fast — there’s enough room for different types of companies to find a relevant angle.
Jessica Rusu, the FCA’s chief data, intelligence, and information officer, put weight on the idea of a trusted environment being the key ingredient here. Without a safe space to test, firms in early discovery or experimental phases basically can’t move. They’re either too cautious or they take risks they shouldn’t. The Sandbox is meant to fix that tension — let companies push harder on AI without the usual regulatory ambiguity hanging over every decision.
The whole setup builds on the Digital Sandbox infrastructure that’s been running for a while now. It’s not a brand-new system from scratch. The FCA added the “Supercharged” layer in 2025, which brought in the compute partnerships and the more structured support model.
The Agentic Academy and the Broader Play
Alongside the Sandbox news, the FCA also launched the Agentic Academy. It’s a 10-week AI program, run with the Centre for Finance, Technology and Entrepreneurship. The idea is to give selected firms specialized AI knowledge — not just access to tools, but actual education on how to use them well. Ten weeks isn’t long, but it’s focused, and it’s paired with everything else the Sandbox offers, so it’s probably more useful than a standalone course would be.
The FCA framed all of this inside a broader pitch about economic growth. Their response to the Prime Minister made the position clear: the FCA wants to support digital innovation without piling on new AI-specific regulations. They’re leaning on existing frameworks rather than building new regulatory architecture around AI. Whether that’s the right call long-term is genuinely unclear, but for now it means firms inside the Sandbox aren’t navigating a new rulebook on top of everything else.
That’s a meaningful design choice. Regulatory uncertainty is one of the biggest friction points for any company trying to move fast on AI in financial services. Removing that uncertainty — or at least reducing it — is probably what’s driving the application surge as much as anything else.
Ubyx partnering with Amazon inside the cohort is worth watching. No details on what that collaboration covers specifically, but Amazon’s cloud and AI infrastructure combined with Sandbox access is a lot of firepower for a single participant.
Full impact data from the first cohort hasn’t been published yet. The FCA hasn’t put out numbers on what the first 132-application round actually produced in terms of deployable products or regulatory learnings. That information is still pending.
Frequently Asked Questions
What AI tools is Anthropic providing to the FCA Supercharged Sandbox?
Anthropic is giving participating firms access to Claude Code and Claude Cowork to support AI development inside the FCA’s controlled testing environment.
How many applications did the FCA’s Supercharged Sandbox second cohort receive?
The FCA received 199 applications for the second cohort, up from 132 in the first round — a 51% increase.
