Anthropic has partnered with the Financial Conduct Authority (FCA) to support the FCA Supercharged Sandbox, providing 21 firms with access to Claude AI models to safely test agent-led payments, fraud detection, and AI governance within a regulated UK environment. The program targets critical industry pain points such as economic crime detection and regulatory compliance automation, following a 51% increase in applications for the second cohort. 21 organisations were selected, including TrueLayer and Scottish Widows, for a 10-week specialist AI training via the new Agentic Academy, after receiving 199 total applications. The infrastructure, supported by NayaOne and NVIDIA, provides accelerated compute power to stress-test large language models against real-world financial data sets. Participants focus on five core pillars, including agent-led commerce and widening financial access for underserved populations. The FCA is utilizing the sandbox to foster growth while relying on existing regulatory frameworks to ensure responsible AI adoption aligned with consumer protection standards.