Bailey Calls for Rigorous AI Testing Before Regulation
Bank of England Governor Andrew Bailey stated that regulating artificial intelligence is “not the right place to start,” emphasizing the need for rigorous testing and safeguards to manage potential risks. Speaking at a financial conference in London, Bailey argued that policymakers should first understand AI’s capabilities and vulnerabilities before imposing binding rules. His remarks come amid growing global debate over how to govern rapidly advancing AI technologies.
Why Regulation Isn’t the First Step
Bailey explained that premature regulation could stifle innovation and fail to address real-world risks. He stressed that AI systems must undergo extensive evaluation to identify failure modes, biases, and unintended consequences. According to the governor, a “test and learn” approach would allow regulators to craft evidence-based policies rather than speculative ones. This perspective aligns with his broader philosophy of proportionate oversight in the financial sector.
The governor’s comments reflect a cautious but forward-looking stance among central bankers. He acknowledged that AI offers significant benefits for productivity and risk management in banking, but warned that unchecked deployment could threaten financial stability. By prioritizing testing, Bailey aims to build a foundation of trust and reliability before any legal framework is enacted. His approach echoes recommendations from industry analysts who advocate for adaptive governance.
AI’s Impact on Financial Stability
AI is increasingly used in high-frequency trading, credit scoring, fraud detection, and customer service, making it integral to modern finance. However, Bailey noted that the complexity of these systems creates new channels for systemic risk. A poorly designed algorithm could amplify market shocks or discriminate against borrowers, undermining public confidence. The Bank of England has been monitoring these developments closely, collaborating with other regulators to share insights.
Recent stress tests conducted by the central bank simulated AI-driven trading scenarios, revealing potential liquidity gaps. These findings underscore the need for robust internal controls and continuous monitoring. Bailey emphasized that financial institutions must invest in explainable AI models to ensure accountability. He also called for international coordination, since AI systems often operate across borders, complicating oversight.
Safeguards and Best Practices
Instead of immediate legislation, Bailey proposed a set of voluntary safeguards that firms should adopt. These include regular audits, bias testing, and clear governance structures for AI deployment. He also recommended that companies maintain human oversight for high-stakes decisions, such as loan approvals or trading strategies. By embedding these practices, the industry can demonstrate responsible innovation while regulators observe and refine their approaches.
The governor’s comments align with the Bank of England’s ongoing work on operational resilience. In recent years, the bank has issued guidance on third-party risk and cyber security, which now extends to AI vendors. Bailey stressed that safeguards should be proportionate to the size and complexity of the institution, avoiding a one-size-fits-all mandate. This flexibility is designed to encourage adoption among smaller banks that may lack extensive AI expertise.
Global Regulatory Landscape
Bailey’s stance contrasts with the European Union’s AI Act, which takes a risk-based approach with binding requirements for high-risk systems. Meanwhile, the United States has favored sector-specific guidance, leaving much to self-regulation. The Bank of England governor suggested that the UK could serve as a bridge between these models, promoting innovation while maintaining safety. However, he cautioned that divergence in rules could lead to regulatory arbitrage, where firms relocate to less stringent jurisdictions.
International forums, such as the Financial Stability Board, are working to harmonize principles for AI in finance. Bailey expressed optimism that consensus could be reached on core issues like transparency and accountability. He noted that the G7’s Hiroshima AI Process has already produced voluntary guidelines, which could inform future binding standards. For now, the priority remains building empirical evidence through rigorous testing and shared learning.
Industry Reaction and Next Steps
Banking industry groups have welcomed Bailey’s pragmatic approach, noting that it gives them time to develop robust AI systems without fear of sudden legal constraints. Some technology companies argue that regulation could hinder their ability to customize AI for niche applications. Consumer advocates, however, express concern that voluntary measures may not adequately protect vulnerable populations. They point to instances of algorithmic bias that have already caused harm.
The Bank of England plans to publish a discussion paper later this year, outlining potential policy options for AI oversight. This document will invite feedback from stakeholders, including financial firms, academics, and civil society. Bailey indicated that any future regulation would be based on this consultation and the results of ongoing testing. He reiterated that the goal is not to slow down progress but to ensure that AI serves society safely and fairly.
As AI continues to evolve, the debate over its governance will intensify. Bailey’s insistence on rigorous testing represents a deliberate attempt to avoid knee-jerk reactions. By focusing on evidence, he hopes to craft rules that are both effective and adaptable. The coming months will reveal whether his approach gains traction among other regulators and policymakers.
