Anthropic joins UK FCA’s AI regulatory sandbox as second cohort launches

Anthropic will provide its Claude AI models to companies participating in the UK Financial Conduct Authority’s (FCA) second Supercharged Sandbox cohort. This initiative is part of the FCA’s efforts to create a regulated environment where financial firms can test AI applications safely. The Claude models available include Claude Code and Claude Cowork, which participants can use to develop solutions for agent-led payments, fraud detection, compliance automation, and AI governance.

The second cohort of the Supercharged Sandbox includes 21 organizations, such as Scottish Widows, Money Advice Trust, and TrueLayer. The FCA reported receiving 199 applications for this cohort, marking a 51% increase over the first group, indicating growing interest from financial companies in experimenting with AI under regulatory oversight. This program complements the FCA’s partnerships with technology providers NVIDIA and NayaOne, which supply the computing infrastructure and development tools necessary for the participants.

Alongside the launch of the second SandBox cohort, the FCA introduced a 10-week Agentic Academy in collaboration with the Centre for Finance, Technology and Entrepreneurship. This academy is intended to assist firms in developing and deploying AI agents specifically for financial services. The FCA distinguishes this Supercharged Sandbox, which targets firms still experimenting with AI, from its AI Live Testing program designed for firms ready to implement AI in live financial markets. The FCA’s initiatives reflect a proactive regulatory approach to AI innovation in finance.