The Financial Conduct Authority has launched a second cohort of firms under its AI Live Testing initiative, enabling selected institutions to deploy artificial intelligence systems in live UK financial markets under direct regulatory oversight.
Firms including Barclays, Experian, Lloyds Banking Group, and UBS are testing AI applications across credit assessment, anti-money laundering controls, and investment support functions.
The initiative does not introduce new AI-specific rules. Instead, it operates within existing regulatory frameworks, indicating that AI-driven activities must be governed, monitored, and controlled in a manner consistent with established obligations relating to consumer protection, financial crime prevention, and operational resilience.
From a compliance perspective, this places immediate emphasis on whether organisations can demonstrate how AI-supported decisions are made, how outputs are monitored in practice, and where accountability sits within existing governance structures.
Firms may therefore need to review documentation, oversight mechanisms, and control processes to ensure that AI deployment can withstand supervisory scrutiny under current regulatory standards.
Regulatory Context and What Is Known
The Financial Conduct Authority confirmed on 21 April 2026 that eight firms have been selected to participate in the second cohort of its AI Live Testing initiative.
The programme enables participating institutions to deploy artificial intelligence applications in live financial environments while working directly with the regulator and its technical partner, Advai.
The initiative is not an enforcement action. It operates as a structured supervisory framework designed to examine how firms manage risk, monitoring, and governance when AI systems are deployed in real-world financial operations, and to assess how existing regulatory expectations apply in practice.
Firms in the second cohort are testing a range of use cases, including credit scoring insights, anti-money laundering detection, Know Your Customer processes, AI-supported investment tools, and agentic payment systems.
These applications reflect the expanding use of advanced AI models—such as agentic systems, small language models, and neurosymbolic approaches—across both consumer-facing and internal operational functions within financial services, alongside broader trends in generative AI adoption in legal practice.
Testing commenced in April 2026 and is expected to continue through to the end of the year.
The FCA has confirmed that it will publish a report on good and poor practices later in 2026, followed by a formal evaluation of the initiative in Q1 2027.
These outputs are intended to provide further clarity on how AI can be deployed safely and responsibly within regulated financial markets.
The FCA has also indicated that it does not intend to introduce new AI-specific regulation at this stage.
Instead, firms are expected to operate within existing regulatory frameworks, particularly those relating to consumer protection, financial crime prevention, market integrity, and operational resilience.
In practical terms, this means that AI is not treated as a separate regulatory category, but must meet the same legal and compliance standards as any system used within regulated financial activities.
From a compliance perspective, the immediate issue is evidential rather than technical. Firms must be able to demonstrate how AI-supported decisions are made, how outputs are monitored in practice, and where accountability sits within existing governance structures.
This places emphasis on documentation, oversight, and control effectiveness, rather than on the underlying technology itself.
Procedural Context, Compliance Obligations and Liability Risk
The AI Live Testing initiative sits within the Financial Conduct Authority’s broader strategy to support innovation while maintaining supervisory oversight of financial markets.
It operates alongside established mechanisms such as the Regulatory Sandbox and Innovation Pathways, both of which have experienced sustained growth in participation from regulated firms.
The FCA has reported a 49% increase in applications to its innovation services, with artificial intelligence emerging as a significant area of activity.
This indicates that AI Live Testing forms part of a wider regulatory response to the accelerated adoption of AI technologies across financial services, rather than an isolated supervisory programme.
Within this context, the initiative reinforces an emerging supervisory expectation that AI systems should be embedded within existing governance and control frameworks rather than treated as standalone technological deployments.
In practice, this requires organisations to establish clear accountability for AI-driven outputs, implement continuous monitoring of system performance, and maintain documented oversight of how decisions are generated, validated, and reviewed within operational processes.
Responsibility for these controls spans multiple levels of the organisation. Boards are expected to maintain oversight of AI governance frameworks and ensure they remain effective in managing risk exposure.
Legal functions must interpret how existing regulatory obligations apply to AI-supported decision-making. Compliance teams are responsible for designing monitoring, reporting, and escalation mechanisms, while risk functions must assess and validate AI-related exposures.
Operational teams must ensure that deployment aligns with regulatory expectations in day-to-day execution.
Although the FCA has not introduced new AI-specific liability rules, the initiative reinforces that existing regulatory obligations continue to apply in full to AI-driven outcomes.
The principal regulatory risk arises where firms are unable to demonstrate how decisions are made, how outputs are monitored in practice, or where accountability sits within governance structures.
A realistic compliance failure scenario would involve an AI system influencing credit decisions without adequate oversight, validation, or documented control.
In such circumstances, regulatory exposure would not stem from the use of AI itself, but from the firm’s inability to evidence that existing governance and control frameworks have been effectively implemented.
Financial, Regulatory and Governance Implications
While no penalties are associated with the AI Live Testing initiative itself, the potential consequences of inadequate governance are significant.
Firms may face increased supervisory scrutiny, operational disruption where artificial intelligence systems require modification or withdrawal, and higher compliance costs associated with strengthening internal controls and oversight mechanisms.
There is also a material reputational dimension, particularly where AI systems influence consumer-facing outcomes such as credit assessment, fraud detection, or financial decision-making processes, reflecting wider trends in motor finance compensation and regulatory liability.
In practice, weaknesses in governance, monitoring, or documentation may constrain a firm’s ability to scale AI deployment or delay implementation timelines due to heightened supervisory engagement.
The Financial Conduct Authority’s approach reflects a broader regulatory shift towards supervision based on observed real-world performance rather than immediate rule-making.
By using live testing environments, the regulator is able to assess how AI systems operate under real operational conditions and identify where existing regulatory frameworks are sufficient, or where further clarification may be required.
This approach also places participating firms within a supervisory development context where practical outcomes may inform future expectations.
The planned publication of good and poor practice guidance later in 2026 will provide a structured benchmark for evaluating AI governance approaches across the wider financial services sector.
Against this backdrop, organisations should treat the initiative as a prompt to review their internal approach to AI governance and control design.
This includes ensuring that accountability for AI systems is clearly defined, that monitoring processes operate continuously rather than at fixed intervals, and that AI-supported decision-making is consistently documented, traceable, and subject to appropriate review within existing control frameworks.
There is also a need to ensure that escalation pathways are clearly defined where AI-driven outcomes are unexpected, adverse, or uncertain, particularly in consumer-facing contexts.
Firms should assess whether existing governance and control arrangements would withstand supervisory scrutiny based on real-world system performance, rather than theoretical design assumptions, as regulatory expectations continue to evolve through observed practice.
Key Compliance Questions
What is the FCA AI Live Testing initiative?
It is a regulatory programme allowing firms to deploy AI systems in live financial environments under FCA supervision to assess risk management, monitoring, and governance.
How does the FCA regulate AI in financial services?
The FCA has indicated that AI systems must comply with existing regulatory frameworks, including those relating to consumer protection, financial crime, and operational resilience.
What compliance risks do AI systems create for firms?
Risk arises where firms cannot demonstrate how AI decisions are made, monitored, or controlled within existing regulatory obligations.
Do firms need new governance structures for AI?
The FCA has not introduced new rules, but firms may need to ensure existing governance frameworks extend effectively to AI-driven processes.
What should organisations do now?
Firms should review governance, monitoring, and accountability for AI systems to ensure alignment with current regulatory expectations.
Key Takeaways and Strategic Implications
AI deployment in financial services is increasingly being assessed in live operational environments under direct regulatory oversight.
Existing regulatory frameworks continue to apply in full, meaning firms must demonstrate effective governance, monitoring, and accountability for AI-driven systems in practice.
Regulatory expectations are therefore being shaped through observed real-world application rather than immediate legislative change, increasing the importance of operational readiness.
Organisations that fail to align with these expectations at an early stage may face heightened supervisory scrutiny, remediation requirements, and increased compliance costs.
The significance of the AI Live Testing programme lies in how these expectations are being formed.
Rather than waiting for new statutory requirements, the Financial Conduct Authority is using real-world evidence to evaluate what constitutes acceptable practice in the deployment of artificial intelligence across regulated financial services.
For organisations, this shifts the nature of compliance risk from a question of whether rules exist to a question of whether systems perform effectively under regulatory scrutiny.
The key issue is no longer the presence of a governance framework in theory, but its demonstrable effectiveness in operational practice.
Where firms are unable to evidence robust governance, monitoring, and accountability, the potential consequences include increased supervisory intervention, constraints on AI deployment, and higher operational and compliance burdens.
Over time, this may also influence competitive positioning and broader regulatory standing within the financial services sector.
Regulatory Timeline and Next Phase
Testing will continue throughout 2026 under the Financial Conduct Authority AI Live Testing initiative, with the regulator expected to publish a report on good and poor practices later in the year. A full evaluation of the programme is scheduled for Q1 2027.
These outputs will provide structured insight into how artificial intelligence is being deployed within live financial environments and how existing regulatory frameworks are applied in practice, positioning this development within Legal Tech & Innovation.
They are intended to support greater consistency in supervisory expectations across financial services firms operating under current FCA rules.
For organisations, this phase represents the transition from experimental deployment to emerging supervisory benchmarking.
Firms should therefore anticipate that findings from the programme will inform how AI governance, monitoring, and accountability are assessed under existing regulatory frameworks, particularly in relation to operational resilience and conduct expectations.