Bank of England Governor urges testing over immediate AI regulation

Andrew Bailey, the Governor of the Bank of England, has argued that establishing a regulatory framework for artificial intelligence is not the appropriate initial step. Instead, he has called for rigorous testing to identify vulnerabilities and develop safeguards that can contain emerging risks. In his first article published on the Substack platform, Bailey acknowledged that the risks associated with AI are real and increasingly significant, while simultaneously emphasising that the benefits of the technology are immense.

Bailey stated that the development of AI should neither be halted nor prohibited. However, he insisted that a system for intervention must be established to define the boundaries within which the technology operates. This perspective comes as the debate over the potential dangers of rapid AI advancement intensifies. Leaders of major AI firms, including Anthropic and OpenAI, have recently advocated for slowing down development and adopting an internationally coordinated approach to assess risks and implement safeguards. OpenAI recently announced that it would not release its latest model due to safety concerns, a move that aligns with calls for greater caution in the sector.

The proposed slowdown has faced opposition from US President Donald Trump, who has stated that the United States is leading the AI race ahead of China and intends to maintain that position. Trump recently hosted a meeting with executives from OpenAI, Anthropic, Nvidia, SpaceX, Meta, and Google. Following the discussion, he reported that the executives had signed a morally binding document intended to serve as a form of protection against AI risks. Under this agreement, the participating companies are responsible for ensuring the safety of their own technologies.

In the United Kingdom, Bailey noted that important work is already underway through the AI Security Institute. He suggested that this work should help form part of a broader set of standards. While acknowledging that the pace of progress must accelerate, Bailey urged that the process should proceed with a degree of humility. He recognised that testing would inevitably lead to failures and unexpected model behaviour, but argued that such outcomes demonstrate the necessity of testing rather than its failure. Bailey clarified that testing should not be viewed as an alternative to future regulation, noting that a more formal regulatory framework might emerge over time. He specifically highlighted the risks associated with frontier AI, describing these highly advanced self-learning systems as potentially becoming a closed loop where the model progressively governs itself without effective intervention.