Bank of England Governor Andrew Bailey says rigorous AI testing should come before formal regulation, while authorities retain the ability to intervene as risks grow. Key Points: Bailey says
Bank of England Governor Andrew Bailey says rigorous AI testing should come before formal regulation, while authorities retain the ability to intervene as risks grow.
- Bailey says AI risks are growing, but policymakers should identify vulnerabilities before building a formal regulatory framework.
- He wants rigorous testing before and after deployment, alongside clear safeguards and credible points for human intervention.
- His comments come as major AI companies and the U.S. government debate voluntary controls, slower development and future regulation.
Bailey said AI risks are “real and increasingly significant,” but policymakers should identify where systems can fail before deciding how formal rules should work. “But regulation is not, in my view, the right place to start,” he wrote.
Bailey argued that society should preserve the ability to set boundaries around advanced AI, using rigorous testing before and after deployment to expose vulnerabilities and check whether safeguards can contain failures. He said AI development should not be prohibited because “the potential benefits are immense.”
The UK has already established the AI Security Institute, which Bailey said is helping build the scientific and technical foundations for AI assurance. “But the pace of progress must accelerate,” he wrote. Bailey added that unexpected model behavior does not show testing has failed, but instead demonstrates why testing is necessary.
Also Read:Ripple Helps Brazil's CSD BR Mirror Fund Ownership On A Public Blockchain
Leading AI companies, including OpenAI and Anthropic, have pushed for stronger safeguards and internationally coordinated risk assessments as concern over increasingly capable systems grows. OpenAI recently said it would not release its latest AI model because of safety concerns.
President Donald Trump has rejected calls to slow U.S. development in the race with China, saying, “whoever wins AI, wins.” After a Sept. 29 White House meeting with major technology executives, Trump said participants signed a “morally binding” AI document that was a “form of protection.”
Under that agreement, participating companies are responsible for the safety of their own technology. Bailey’s approach differs in emphasis because he wants testing to establish where intervention is necessary before policymakers build a more formal regulatory structure around advanced systems.
The debate has accelerated in recent weeks as AI developers, governments and financial authorities have confronted faster model development and more difficult questions about oversight. Bailey’s latest intervention follows industry calls for coordinated safeguards, OpenAI’s decision to withhold a model and the White House push for voluntary company-led controls.
Read Next:Altcoins Take 41% Of Futures Open Interest, Spot Volume Hits 4X Bitcoin
