Bank of England Chief: Begin AI with Rigorous Testing, Not Regulation

Governor of Bank of England: AI Regulation Should Follow Thorough Testing

Andrew Bailey, the Governor of the Bank of England, has expressed his views on managing the risks associated with artificial intelligence (AI). Contrary to immediately jumping into regulation, Bailey advocates for in-depth testing of AI systems to uncover vulnerabilities and develop effective safeguards.

In his debut article on Substack, Bailey emphasized the significant risks AI poses, which are becoming more pronounced with its rapid development. Instead of hindering AI advancements, which he acknowledges hold immense benefits, Bailey believes in establishing a robust framework for intervention and operating boundaries.

The global dialogue around AI risks has gained traction recently, with prominent figures in the technology sector calling for a regulated approach to mitigate potential dangers. Executives from top AI companies such as OpenAI and Anthropic have argued for a slowdown in AI development to evaluate and address associated risks through a coordinated international effort. However, this perspective is not universally shared.

Former U.S. President Donald Trump, for instance, has dismissed the notion of slowing down AI progress, emphasizing that the U.S. maintains an edge in competing with China in AI innovation. Yet, following discussions with AI industry leaders, Trump announced a mutual agreement aimed at protecting against AI’s potential threats.

In the UK, proactive efforts are underway to advance AI testing. The establishment of the AI Security Institute is a significant stride in ensuring comprehensive and rigorous testing becomes a part of standard practice, as highlighted by Bailey. While he calls for a swift progression in these efforts, Bailey stresses the importance of maintaining humility throughout the process. Recognizing that failures and unexpected model behaviors will occur, he views these challenges not as failures of testing but as validations of its necessity.

Bailey insists that initial focus on testing does not negate the future need for regulatory structures. As AI evolves, a more formalized approach to regulation may emerge. However, he firmly believes that regulatory measures are not the starting point for addressing AI’s complexities.

Photo by Ivan Bandura on Unsplash