The UK’s finance watchdog is handing firms Claude to build AI agents that move money
Key Points
- The FCA has partnered with Anthropic to give the second cohort of its Supercharged Sandbox access to Claude, including Claude Code and Claude Cowork.
- 21 organisations made the cut, including Scottish Widows, Money Advice Trust and TrueLayer.
- Applications jumped 51% to 199, up from 132 for the first cohort.
- Firms will test agent-led payments, fraud detection, AI governance and tools for underserved consumers.
- The FCA has ruled out new AI regulation, relying on existing frameworks instead.
The Financial Conduct Authority will give 21 firms, including Scottish Widows, Money Advice Trust and TrueLayer, access to Anthropic’s Claude to build and test AI systems inside its Supercharged Sandbox.
The regulator announced the partnership on Wednesday (22 July) alongside the second cohort of the programme, which gives firms a controlled environment to experiment with advanced AI before deploying it in live financial services.
Anthropic will provide participants with access to Claude, including its agentic coding tool Claude Code and its knowledge work product Claude Cowork, to accelerate development during the programme.
Demand for a place in the sandbox surged ahead of the second intake. The FCA received 199 applications, a 51% increase on the 132 submitted for the first cohort, before selecting the final 21.
What the firms will build
The second cohort will test a set of use cases that reach deep into how money moves through the UK financial system.
These include safer agent-led payments and commerce, where AI agents transact on behalf of consumers and businesses, alongside more effective fraud and economic crime detection, stronger AI governance and accountability frameworks, and tools to widen access to financial services for vulnerable and underserved consumers.
Firms will also test solutions to streamline compliance and business automation, an area where financial institutions spend heavily on manual processes.
Jessica Rusu, Chief Data, Intelligence and Information Officer at the FCA, said the level of interest demonstrates demand for trusted environments where firms can experiment safely and responsibly.
“With the support of Anthropic, participants will benefit from the technology they need to accelerate innovation,” Rusu said.
“This is central to our commitment to supporting economic growth, enabling firms to make the most of technological advances while maintaining the UK’s position at the forefront of responsible AI adoption and innovation.”
A regulator betting on light-touch AI rules
The Supercharged Sandbox builds on the FCA’s existing Digital Sandbox infrastructure, which fintech platform NayaOne provides, and a collaboration with NVIDIA that gave the first cohort access to accelerated computing infrastructure and NVIDIA AI Enterprise software. Firms in the second group keep access to those capabilities.
The regulator has also launched the Agentic Academy, a 10-week specialist AI programme for selected firms, delivered with the Centre for Finance, Technology and Entrepreneurship.
The programme sits within a deliberately light-touch approach to AI oversight. In its response to the Prime Minister’s growth letter, the FCA committed to avoiding additional regulations for AI, relying instead on existing frameworks to govern how firms deploy the technology.
That positions the UK differently from the EU, where the AI Act imposes specific obligations on high-risk AI systems in financial services, and hands firms in the sandbox an unusual advantage: frontier AI tools, regulator support, and no new rulebook to navigate.
