The FCA has named 21 organisations for the second cohort of its Supercharged Sandbox, its controlled environment for experimenting with advanced AI, after applications rose 51% to 199. Anthropic joins existing partners NayaOne and NVIDIA, giving participants access to Claude models, and a new 10-week Agentic Academy launches alongside.
TL;DR · LAST REVIEWED 22 July 2026
- The FCA announced on 22 July 2026 that 21 organisations will join the second group of its Supercharged Sandbox, including Scottish Widows, Money Advice Trust and TrueLayer.
- Applications rose 51% on the first cohort, with 199 firms applying against 132 for the first group.
- Anthropic will support the second cohort, providing participants with access to its Claude models, including Claude Code and Claude Cowork, joining existing sandbox partners NayaOne and NVIDIA.
- Use cases for the second group include safer agent-led payments, fraud and economic crime detection, AI governance, widening access for vulnerable and underserved consumers, and compliance automation.
- The FCA 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 regulator has said it will avoid additional regulations for AI, relying on existing frameworks such as the Consumer Duty and the Senior Managers regime.
FCA Supercharged Sandbox second cohort at a glance
| Cohort size | 21 organisations | Second group, announced 22 July 2026 |
| Applications | 199, up 51% on the first group's 132 | FCA figures |
| Named participants | Scottish Widows, Money Advice Trust, TrueLayer and 18 others | Full list in the FCA's notes to editors |
| AI model access | Claude, including Claude Code and Claude Cowork | Provided by Anthropic for the second cohort |
| Compute and infrastructure | NVIDIA accelerated computing; NayaOne Digital Sandbox | Continuing from the first cohort |
| Agentic Academy | 10-week specialist AI programme | Delivered by the FCA and CFTE for selected firms |
KEY FACTS
- Second cohort: 21 organisations, including Scottish Widows, Money Advice Trust, TrueLayer, Sardine AI, Ubyx (partnering with Amazon) and FSCom
- Demand: 199 applications, up 51% from 132 for the first cohort
- Technology partners: Anthropic (Claude models including Claude Code and Claude Cowork), NVIDIA (accelerated computing and AI Enterprise software), NayaOne (Digital Sandbox infrastructure)
- New: Agentic Academy, a 10-week specialist AI programme delivered by the FCA and CFTE for selected firms
- Policy stance: the FCA has said it will not create additional AI-specific regulations, applying existing rules to AI instead
What has the FCA announced
The FCA has named the 21 organisations joining the second cohort of its Supercharged Sandbox and confirmed that Anthropic will support the group, providing access to its Claude models alongside the computing infrastructure already supplied by NVIDIA and NayaOne.
The Supercharged Sandbox is the FCA's controlled environment in which financial services firms can experiment with advanced AI safely, aimed at firms in the discovery and experimentation phase of AI adoption. For the second cohort, announced on 22 July 2026, Anthropic becomes the latest technology partner, giving participating firms access to its Claude models, including the Claude Code developer tool and Claude Cowork, to speed up development work during the programme. That sits on top of the arrangements carried over from the first cohort: accelerated computing infrastructure and AI Enterprise software from NVIDIA, and the underlying Digital Sandbox platform provided by NayaOne. Jessica Rusu, the FCA's chief data, intelligence and information officer, said the level of interest showed the demand for trusted environments where firms can experiment safely and responsibly, and framed the programme as part of the regulator's commitment to supporting economic growth while keeping the UK at the forefront of responsible AI adoption.
Which firms are in the second cohort
21 organisations join the second group, including household names Scottish Widows and Money Advice Trust and open banking firm TrueLayer, selected from 199 applications, a 51% increase on the 132 received for the first cohort.
The best known names in the second group span very different corners of financial services: Scottish Widows, the pensions and insurance brand; Money Advice Trust, the debt advice charity; and TrueLayer, the open banking payments firm. The FCA's notes to editors list the remaining participants as calQrisk, Merx Digital Solutions (SmartDrops), Aegis Trace, Sardine AI, Zquas, Trustie Labs, Welleness, IntelXview, RMI Agentic, Ubyx (partnering with Amazon), Deepflow, FSCom, GAI Labs, Condukt, Kaption and Relace. The mix, from a major pensions provider and a debt charity through to early stage AI startups, reflects the sandbox's positioning as a discovery-phase programme rather than one for production deployments. The selection followed what the FCA describes as strong demand: 199 applications against 132 for the first iteration, a rise of 51%, which the regulator reads as evidence of appetite among firms for structured, supervised environments in which to test AI rather than experimenting entirely on their own.
What will the firms be testing
The second cohort's use cases cover safer agent-led payments and commerce, more effective fraud and economic crime detection, stronger AI governance and accountability, wider access for vulnerable and underserved consumers, and streamlined compliance and business automation.
Two of the five themes carry direct consumer relevance. The first is agent-led payments and commerce: AI agents that can initiate transactions on a customer's behalf raise obvious questions about authorisation, fraud and liability, and the sandbox work is aimed at making that model safer before it reaches the mainstream. The second is access: some participants will test solutions designed to widen access to financial services for vulnerable and underserved consumers, an area where the Money Advice Trust's presence in the cohort is notable. The remaining themes are more industry-facing but still land on consumers indirectly: fraud and economic crime detection shapes how quickly scams are caught, AI governance and accountability determines who answers when an AI system gets a decision wrong, and compliance automation affects the cost base of regulated firms. The FCA has not said which firm is testing which use case, and sandbox participation is explicitly experimental: none of the work implies that any resulting product is approved or endorsed by the regulator.
What is the Supercharged Sandbox
Launched in 2025, the Supercharged Sandbox is the FCA's safe testing environment for firms in the discovery and experimentation phase of AI, built on the NayaOne Digital Sandbox infrastructure with advanced compute supplied through the NVIDIA collaboration.
The FCA announced the Supercharged Sandbox in 2025 as an extension of its established Digital Sandbox, the testing platform run on NayaOne infrastructure that gives firms synthetic data and development tools without touching live customer systems. The supercharged version adds the computing power that modern AI development demands: through the NVIDIA collaboration, first-cohort firms had access to accelerated computing infrastructure and NVIDIA AI Enterprise software, and the second group keeps those capabilities alongside the new Claude model access. The programme sits within the FCA's broader innovation function, which has run regulatory sandboxes for fintech firms since 2016, and is aimed specifically at the earliest stages of AI work, where firms are still establishing what is feasible and safe rather than seeking authorisation for a finished product. Alongside the second cohort, the FCA has launched the Agentic Academy, a 10-week specialist AI programme for selected firms delivered with the Centre for Finance, Technology and Entrepreneurship, focused on the agentic AI systems that featured heavily in the cohort's use cases.
What does this signal about UK AI regulation in finance
The FCA's stated approach is to avoid additional AI-specific regulations and apply existing frameworks instead, using the sandbox to shape how current rules meet AI in practice while positioning AI adoption as part of the UK's growth agenda.
The sandbox expansion lands within a deliberate regulatory posture. In its response to the Prime Minister's letter on supporting growth, the FCA said it would avoid creating additional regulations for AI, relying instead on existing frameworks, and its published approach explains how current rules, from the Consumer Duty to the Senior Managers and Certification Regime, already apply to firms' use of AI. That places the UK on a different path from jurisdictions building standalone AI rulebooks: rather than new law, the FCA's bet is that outcome-based rules plus supervised experimentation will keep pace with the technology. The sandbox is the practical half of that bet, giving the regulator early sight of how firms actually deploy AI, in agent payments, fraud detection and consumer access, before those systems reach the market at scale. For consumers, the near-term effect is indirect but real: the use cases being tested shape how the next generation of banking, payments and advice tools will work, and the governance and accountability strand goes to the question that matters most when AI enters financial decisions, which is who is responsible when it goes wrong.
RELATED GUIDES
DISCLAIMER
This article is for general information only and does not constitute financial advice. Participation in an FCA sandbox is experimental and does not amount to FCA approval or endorsement of any firm, product or technology mentioned.
Frequently asked questions
What is the FCA Supercharged Sandbox
A controlled environment launched by the FCA in 2025 where financial services firms can safely experiment with advanced AI. It builds on the Digital Sandbox infrastructure provided by NayaOne and offers advanced computing power for AI development.
Which firms are in the second Supercharged Sandbox cohort
21 organisations, including Scottish Widows, Money Advice Trust, TrueLayer, calQrisk, Sardine AI, Aegis Trace, Ubyx (partnering with Amazon), FSCom and others listed in the FCA's notes to editors, selected from 199 applications.
What is Anthropic providing to sandbox firms
Access to its Claude AI models, including Claude Code and Claude Cowork, to help participants speed up their development work. Anthropic joins existing sandbox partners NayaOne and NVIDIA.
What is the FCA Agentic Academy
A new 10-week specialist AI programme for selected firms, delivered by the FCA with the Centre for Finance, Technology and Entrepreneurship, announced alongside the second sandbox cohort.
Is the FCA creating new rules for AI
No. The FCA has said it will avoid additional regulations for AI, relying on existing frameworks instead, and has published guidance on how its current rules apply to firms' use of AI.
SOURCES
- FCA press release, Anthropic to support FCA's Supercharged Sandbox – accessed 22 July 2026
- FCA, Supercharged Sandbox – accessed 22 July 2026
- FCA, how FCA rules apply to AI – accessed 22 July 2026
