Anthropic joins FCA AI sandbox for financial services testing
The UK regulator’s second Supercharged Sandbox drew 199 applications, with 21 firms selected to test AI uses in payments, fraud and compliance.
By Rafael Ortiz · Fintech Correspondent
· 3 min read
Anthropic will support the UK Financial Conduct Authority’s Supercharged Sandbox, giving selected financial services firms access to Claude Code and Claude Cowork as they test artificial intelligence applications. The second cohort attracted 199 applications, with 21 firms chosen to take part, according to Finextra.
The sandbox is a controlled environment designed to let firms experiment with advanced AI tools under supervision. Such programmes give participants a setting in which they can test prototypes, data workflows and operational controls before deciding whether a system is suitable for wider deployment.
Finextra reported that Anthropic is joining Naya One and Nvidia as a technology supporter for the programme. Participating firms will use the second cohort to work on applications covering agent-led payments, fraud detection, AI governance and accountability, compliance processes and services intended to broaden access for financially excluded customers.
Participants and areas of testing
The named firms selected for the cohort include calQrisk, Merx Digital Solutions Ltd, through SmartDrops, Aegis Trace, Sardine AI Corp, Zquas, Trustie Labs, Welleness, IntelXview Ltd, RMI Agentic, Ubyx in partnership with Amazon, Deepflow, FSCom, GAI Labs, Condukt, Kaption and Relace. Finextra said 21 firms were selected in total.
The use cases reflect several pressure points for banks, payment firms and market infrastructure providers. Agent-led payments involve software systems that can initiate or manage payment-related actions on behalf of users within defined rules. Fraud detection tools typically use data patterns to identify suspicious behaviour, while AI governance projects focus on oversight, accountability and controls around model use.
Compliance applications are expected to test whether AI can reduce manual work in regulated processes. The FCA has also included projects aimed at widening financial access, an area where regulators often scrutinise both customer outcomes and the risk that automated systems could reinforce exclusion if controls are weak.
Regulatory support for AI experimentation
Jessica Rusu, the FCA’s chief data, intelligence and information officer, said demand for the programme showed firms wanted trusted spaces to test technology safely and responsibly. “With the support of Anthropic, participants will benefit from the technology they need to accelerate innovation,” Rusu said, according to Finextra.
The FCA has also launched an Agentic Academy, a 10-week specialist AI programme for selected firms. Finextra reported that the programme is delivered by the FCA and the Centre for Finance, Technology and Entrepreneurship.
The regulator’s sandbox approach places experimentation inside a defined framework rather than leaving firms to test new AI systems entirely in live settings. For financial companies, that structure can help clarify how models interact with existing rules on operational resilience, consumer protection, data use and accountability.
The announcement adds Anthropic to a group of firms providing technical support to the FCA initiative as policymakers assess how AI may alter payments, compliance and customer services. The FCA has not been reported as providing commercial projections for the sandbox or details of any eventual authorisations linked to the tests.
This story draws on original reporting from Finextra Research.
