Andrew Bailey, Governor of the Bank of England, has argued that imposing immediate regulation on artificial intelligence is not the appropriate initial approach. In his first-ever Substack article, Bailey emphasized that the priority must be rigorous testing to uncover vulnerabilities and create safeguards that contain potential risks.
Bailey acknowledged that the dangers posed by AI are “real and increasingly significant,” yet he cautioned against halting or prohibiting its development. Instead, he stressed the immense benefits of the technology and called for a system of intervention to establish operational boundaries.
This stance comes amid intensifying global debate regarding the rapid advancement of AI. Executives from major firms, including Anthropic and OpenAI, have recently advocated for a slowdown in development and an internationally coordinated strategy to assess risks. OpenAI, for instance, announced it would delay the release of its latest model due to safety concerns.
In contrast, US President Donald Trump has rejected calls for a development slowdown, asserting that the United States must maintain its lead over China. Following a meeting with technology leaders from companies such as OpenAI, Nvidia, Meta, Google, and SpaceX, Trump stated that executives had signed a “morally binding” document outlining their responsibility for the safety of their own technology.
Turning to the UK’s approach, Bailey highlighted the work of the AI Security Institute, which is developing standards through testing. He insisted that while the pace of progress must accelerate, the sector should proceed with humility. Bailey noted that unexpected model behavior during testing is not a sign of failure but rather proof of why such testing is essential.
Although he suggested that a formal regulatory framework may eventually emerge, Bailey maintained that regulation should not be the starting point. He specifically warned about “frontier AI”—highly advanced, self-learning systems that, without effective intervention mechanisms, could become a self-governing closed loop.
I agree with testing, but what about liability when things go wrong? The article doesn’t mention that gap.
Does anyone else think calling it a ‘morally binding’ document is just corporate PR without legal teeth?
Interesting that Trump’s approach is completely opposite. The UK seems to be walking a tighter rope here.
But who defines the safety boundaries? Without regulation, aren’t we just handing power to Big Tech?
Finally, a sensible voice in this debate. Testing before rules actually makes sense for such fast-moving tech.