FCA’s Supercharged Sandbox Welcomes Anthropic 2026 Cohort

Table of Contents

• 1. FCA’s Supercharged Sandbox expands AI innovation opportunities
• 2. What is the focus of the FCA’s second Supercharged Sandbox cohort?
• 3. How many applications were received for the second cohort of the Supercharged Sandbox?
• 4. Which organizations were selected for the second cohort of the Supercharged Sandbox?
• 5. What support will Anthropic provide to the participants of the second cohort?
• 6. What are the objectives of the FCA’s Supercharged Sandbox?

FCA’s Supercharged Sandbox expands AI innovation opportunities

FCA Sandbox AI Cohort Update
The FCA’s Supercharged Sandbox is a cohort-based programme that lets firms trial advanced AI in a controlled setting, using shared infrastructure and tooling. For the second cohort (announced July 2026), two things stand out: applications rose to 199 (up 51% from 132), and Anthropic is now providing access to Claude tools to help teams build and test faster.

• The FCA’s second Supercharged Sandbox cohort will test AI use cases spanning agent-led payments and commerce, fraud detection, financial inclusion, and governance.
• Interest is rising: the regulator received 199 applications, up 51% from the first cohort’s 132.
• A total of 21 organisations were selected, including TrueLayer, Scottish Widows and the Money Advice Trust.
• Anthropic will provide access to Claude, including Claude Code and Claude Cowork, to help participants move faster.

What is the focus of the FCA’s second Supercharged Sandbox cohort?

MOVE–PROTECT–INCLUDE–GOVERN–AUTOMATE
A quick way to remember the cohort’s focus is “MOVE–PROTECT–INCLUDE–GOVERN–AUTOMATE”:

• MOVE: agent-led payments and commerce (how transactions are initiated and managed)
• PROTECT: fraud detection and economic crime (how suspicious activity is identified and investigated)
• INCLUDE: financial inclusion (how vulnerable and underserved consumers are supported)
• GOVERN: AI governance and accountability (how oversight, controls, and responsibility are maintained)
• AUTOMATE: compliance and business automation (how regulated processes are streamlined)

The UK’s Financial Conduct Authority (FCA) says the second cohort of its Supercharged Sandbox will concentrate on AI-enabled use cases that sit at the heart of modern financial services: how money moves, how risk is detected, and how consumers—especially those who struggle most—are served.

A major theme is agent-led payments and commerce. In practice, that means exploring scenarios where AI systems act with a degree of autonomy to help initiate or manage transactions, potentially reshaping customer journeys in payments and purchasing. The FCA’s framing is not simply about speed or convenience; it is explicitly about making these flows safer, reflecting the regulator’s concern that more capable AI can also amplify operational and consumer risks if not properly governed.

The cohort will also test tools aimed at detecting fraud and economic crime more effectively. Fraud detection is already a mature application area for machine learning in finance, but the FCA’s emphasis suggests a push toward more advanced AI techniques and workflows—potentially including systems that can reason over complex patterns, triage alerts, or support investigations—while still operating within a controlled environment.

A third pillar is financial inclusion, with projects intended to widen access to financial services for vulnerable and underserved consumers. That focus matters because AI can either reduce friction—through better assistance and automation—or inadvertently exclude people if models fail to handle edge cases, accessibility needs, or non-standard financial histories.

Finally, the FCA highlights AI governance and accountability, alongside streamlining compliance and business automation. Taken together, the second cohort is positioned as both a technology sprint and a governance exercise: proving what advanced AI can do in finance, and demonstrating how it can be deployed responsibly.

How many applications were received for the second cohort of the Supercharged Sandbox?

The FCA reports it received 199 applications for the second Supercharged Sandbox cohort—an increase it describes as 51% higher than the first cohort.

The jump to nearly 200 applications in the second round is a clear signal of demand from firms that want a structured way to experiment with advanced AI while staying close to regulatory expectations.

That demand is notable for two reasons. First, it suggests that AI experimentation in financial services is no longer limited to a handful of early adopters; instead, a broad set of organisations appear to be looking for credible pathways from prototype to production. Second, it underscores why the FCA is leaning into “trusted environments,” where firms can test systems that may be difficult to validate through traditional internal governance alone—particularly when the use cases involve autonomous or semi-autonomous “agentic” behaviours.

Jessica Rusu, the FCA’s chief data, intelligence and information officer, linked the strong interest directly to the need for safe experimentation:

“The high level of interest in the Supercharged Sandbox demonstrates the demand for trusted environments where firms can experiment safely and responsibly.”
Jessica Rusu, chief data, intelligence and information officer, FCA

The FCA’s message is that the application numbers are not just a popularity metric—they are evidence that the market wants clearer guardrails and better tooling for responsible AI development in finance.

Which organizations were selected for the second cohort of the Supercharged Sandbox?

The FCA says 21 organisations have been chosen for the second Supercharged Sandbox cohort. Among those named publicly are a mix that hints at the breadth of interest, from fintech infrastructure to established financial services and consumer-focused organisations.

Beyond those examples, the FCA’s description of the cohort’s work suggests participants will be building and testing solutions across several categories:

• Safer agent-led payments and commerce, where AI systems may help initiate or manage transactions.
• Fraud and economic crime detection, aiming to improve effectiveness in identifying suspicious activity.
• AI governance and accountability, focusing on controls, oversight, and responsible deployment.
• Financial inclusion, with an emphasis on vulnerable and underserved consumers.
• Compliance and business automation, streamlining internal processes under regulatory expectations.

The selection of 21 organisations also reflects a competitive funnel: 199 applicants for 21 places implies the FCA is curating a relatively small group to work intensively within the sandbox environment.

The infrastructure underpinning this work is also part of the story. The FCA’s existing digital sandbox infrastructure—on which the Supercharged Sandbox is built—is provided by NayaOne, reinforcing that the programme is not a one-off experiment but an extension of an established platform.

Karan Jain, founder and CEO of NayaOne, pointed to the application volume as proof of the programme’s relevance:

“Jessica Rusu called it demand for trusted environments, and that’s exactly what the numbers show, nearly 200 applications to experiment safely with leading agentic AI use cases. It’s fantastic to see Anthropic join the journey and support this cohort.”
Karan Jain, founder and CEO, NayaOne

While the FCA has not positioned the cohort as representative of the entire market, the named participants and the use-case spread indicate an effort to cover both consumer-facing and operational applications—where the benefits and risks of advanced AI can be tested under supervision.

What support will Anthropic provide to the participants of the second cohort?

Operationalizing Anthropic Tools in Sandboxes
How Anthropic’s tools map to day-to-day sandbox work:

• Claude → product, risk, and ops teams → draft and iterate customer journeys, policies, and test scenarios faster
• Claude Code → engineering teams → speed up prototyping, code review, and debugging during rapid experiments
• Claude Cowork → cross-functional teams → coordinate workflows (e.g., compliance automation pilots) and reduce handoff friction

Practical checkpoints teams typically need in a regulated sandbox setting:

• Define what the AI is allowed to do vs. what must stay human-approved (especially for agent-led payments)
• Keep an auditable record of prompts/outputs and decision points during testing
• Test failure modes (conflicting instructions, missing data, edge-case customers) before expanding scope

Anthropic’s role in the second cohort is direct and practical: the company will provide participants with access to Claude, including Claude Code and Claude Cowork, with the explicit goal of helping firms “speed up their development work.”

This matters because the Supercharged Sandbox is designed for experimentation with advanced AI in a controlled environment. Access to a frontier model and associated tools can reduce the time it takes to prototype, test, and iterate—especially for teams building AI-enabled workflows that touch payments, fraud detection, compliance, or customer support.

The FCA’s announcement frames Anthropic’s contribution as enabling capability, not just branding. In other words, Anthropic is not merely endorsing the programme; it is supplying the AI tooling that participants will use to build and evaluate their solutions.

Jessica Rusu tied that support to the FCA’s broader economic and innovation agenda:

“With the support of Anthropic, participants will benefit from the technology they need to accelerate innovation. 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.”
Jessica Rusu, chief data, intelligence and information officer, FCA

Anthropic’s tools arrive in a cohort focused on “agentic” use cases—where AI may take more initiative in workflows. In that context, developer-oriented capabilities like Claude Code and collaborative tooling like Claude Cowork are positioned as accelerators for building, testing, and refining systems that must meet both product goals and governance expectations.

What are the objectives of the FCA’s Supercharged Sandbox?

FCA Cohort Testing Priorities
What the FCA is trying to achieve (as reflected in the cohort’s stated testing goals):

• 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
• Streamlined compliance and business automation

A simple way to use this list during the cohort: for each pilot, teams should be able to point to which box(es) it advances—and what evidence from testing supports that.

The FCA launched the Supercharged Sandbox to help firms experiment with advanced AI in a controlled environment—a deliberate attempt to balance innovation with safety, accountability, and consumer protection.

At its core, the programme is about creating a place where companies can test AI systems that may be difficult to validate in live markets without introducing unacceptable risk. That includes AI used in sensitive domains such as payments, fraud detection, and compliance—areas where errors can lead to financial loss, consumer harm, or regulatory breaches.

The second cohort’s stated testing goals provide a clear window into the programme’s objectives. The FCA says participants will test solutions designed to:

• Enable safer agent-led payments and commerce
• Detect fraud and economic crime more effectively
• Strengthen AI governance and accountability
• Widen access to financial services for vulnerable and underserved consumers
• Streamline compliance and business automation

These aims show the FCA is not treating AI as a single “use case.” Instead, it is treating AI as a capability that cuts across the financial system—touching customer journeys, internal controls, and market integrity.

The programme also builds on the FCA’s existing digital sandbox infrastructure, provided by NayaOne. That continuity suggests the regulator is institutionalising sandbox-based experimentation as part of its innovation toolkit, rather than running isolated pilots.

The surge in applications—from 132 in the first cohort to 199 in the second—also reinforces the FCA’s premise: many firms want to innovate with advanced AI, but they want to do so with clearer guardrails and a credible environment for testing. The Supercharged Sandbox is the FCA’s answer to that demand.

This article reflects publicly available information as of late July 2026. Cohort membership, tooling access, and programme details may change as the FCA and partners publish updates. Any organisation “type” labels are high-level descriptors for reader orientation and should not be treated as formal regulatory classifications.