The Financial Conduct Authority (FCA) has expanded its artificial intelligence (AI) testing initiative, welcoming Barclays, Experian and UBS into the second cohort of firms that will trial AI applications in a controlled, real-world environment. These companies join earlier participants such as Lloyds Banking Group, NatWest and Monzo, all benefiting from tailored regulatory support and oversight as they push the boundaries of AI in financial services.
The initiative, which the FCA describes as a safe space for innovation, is designed to help firms that have already made significant progress in AI development and are ready to deploy their solutions in live markets. Participants receive guidance from the regulator and its technical partner, Advai, to navigate key challenges around risk management, monitoring and responsible deployment. The goal is to foster safe and responsible AI adoption that benefits both consumers and market integrity.
Second Cohort and Use Cases
The second group of firms will test a variety of customer-facing and business-to-business applications. Consumer-focused use cases include AI-driven targeted support for investment decisions and credit score insights that help individuals better understand their financial health. On the business side, the cohort will explore agentic payments—where AI agents autonomously initiate and process transactions—and advanced money laundering detection systems that can identify suspicious patterns more efficiently than traditional rule-based methods.
Jessica Rusu, chief data, information and intelligence officer at the FCA, emphasised the collaborative nature of the programme: “We’re continuing to collaborate with firms to support the safe and responsible development of AI in UK financial markets.” She added that the tailored support reflects the regulator’s commitment to keeping pace with rapid AI advancements and demonstrates “how regulators and industry can work together to harness innovation responsibly.”
The FCA noted that participating banks are experimenting with a wide range of technologies, from agentic AI—where autonomous systems make decisions and execute actions—to small language models (SLMs) that are more efficient than large models, and emerging techniques such as neurosymbolic AI, which combines neural networks with symbolic reasoning to improve accuracy and explainability.
Regulatory Oversight and Reporting
Throughout the testing phase, the FCA provides ongoing support and oversight, ensuring that firms adhere to regulatory standards while exploring new capabilities. The regulator plans to publish a report later this year that will highlight both good and poor practices observed during the initiative, offering valuable lessons for the wider financial industry. This transparency is intended to help other organisations avoid common pitfalls and adopt best practices when integrating AI into their operations.
The initiative is part of the FCA’s broader Digital Sandbox, which has been running since 2020 to support fintech innovation. Over time, the sandbox has evolved to address specific challenges around data sharing, digital identity and, increasingly, artificial intelligence. The AI-focused cohort model was introduced in 2025 with the first group of seven firms, and this second cohort expands the programme to a total of ten participants.
Background: The Need for Proactive AI Regulation
Despite the FCA’s efforts, UK financial services regulators have faced sharp criticism from Parliament. Earlier this year, the Treasury Committee condemned regulators for taking what it perceived as a “wait-and-see” approach to AI regulation. In a report, MPs warned that “the UK public and the country’s finance system are exposed to potential serious harm because regulators in the financial sector are not doing enough.” The committee urged regulators to become more proactive in identifying and mitigating risks associated with AI, particularly as its adoption accelerates across banking, insurance and payments.
In response, Sarah Breeden, deputy governor for financial stability at the Bank of England, defended the central bank’s actions. “We share the view that AI has broad, complex and likely long-term implications for how the UK financial system serves the real economy,” she said. “However, we do not agree with the characterisation that the bank is taking a ‘wait-and-see’ approach to the use of AI in financial services.” She highlighted that the Bank has invested heavily in analysing both current and future risks from AI, including its impact on financial stability and operational resilience.
The debate intensified following recent developments with Anthropic’s latest AI model, Mythos, which unearthed decades-old software vulnerabilities in legacy banking systems. In response, major UK banks entered discussions with regulators, finance authorities and national security organisations to assess the risks and determine appropriate mitigation strategies. Meg Hillier, MP and chair of the Treasury Committee, commented: “Recent developments in the world of AI, such as Anthropic’s Project Mythos, show us how fast this transformative technology is moving. It has never been more important that those responsible for maintaining the UK’s financial stability take a proactive approach to understanding and mitigating the risks AI may pose to our financial system.”
Detailed Analysis of Technologies Being Tested
The second cohort’s experiments cover a broad spectrum of AI capabilities. Agentic AI systems, for example, can autonomously manage payment flows, detect anomalies and even negotiate transactions without human intervention. While this promises efficiency gains, it also raises concerns about accountability, error propagation and potential misuse. The FCA’s sandbox allows firms to trial these systems under close supervision, with real-time monitoring and predefined guardrails.
Small language models (SLMs) are gaining traction because they require fewer computational resources and can be fine-tuned for specific financial tasks, such as summarising regulatory documents or generating personalised customer advice. Unlike large general-purpose models, SLMs can operate on-premises, reducing data privacy risks. The FCA’s oversight ensures that these models are trained on appropriate datasets and that their outputs are explainable.
Neurosymbolic AI represents a nascent but promising field that merges the pattern recognition strengths of neural networks with the rule-based reasoning of symbolic AI. This hybrid approach could enhance fraud detection by combining statistical learning with logical inference, making it easier to audit and justify decisions. The FCA’s sandbox provides a unique opportunity to test this technology in a regulated environment where explainability is paramount.
Implications for the Financial Sector
The expansion of the FCA’s AI testing initiative signals that UK regulators are attempting to strike a balance between fostering innovation and safeguarding consumers. By allowing firms to experiment in a controlled setting, the FCA can gather empirical data on what works and what doesn’t, informing future policy and potential rulemaking. The initiative also serves as a model for other jurisdictions looking to regulate AI in finance without stifling progress.
However, the persistent criticism from Parliament suggests that some stakeholders believe the pace of regulation is still too slow. As AI models become more powerful and autonomous, the potential for systemic risk grows. The Treasury Committee has called for mandatory stress testing of AI systems, regular audits and stronger enforcement powers for regulators. The FCA’s upcoming report on good and poor practices will likely influence these debates, providing concrete examples that could shape the next generation of AI governance.
For the firms involved, the benefits of participating in the sandbox extend beyond regulatory guidance. They gain early-mover advantage, access to expert feedback and the ability to shape industry standards. Barclays, Experian and UBS will likely use the insights gained to refine their AI strategies and potentially commercialise new services once the testing phase concludes.
As the initiative progresses, all eyes will be on the FCA’s published findings. If the sandbox demonstrates that responsible AI deployment is achievable without major incidents, it could pave the way for broader adoption across the UK financial sector. Conversely, any failures or near-misses will reinforce the need for stricter safeguards. The coming months will be critical in determining whether the UK’s approach to AI regulation is a model of innovation-friendly governance or a cautionary tale of reactive oversight.
Source: ComputerWeekly.com News