Enhanced AI testing trumps regulation, says Bank of England governor

In an article accompanying the Bank of England's Financial Policy Committee quarterly meeting report, Governor Andrew Bailey has emphasized the importance of rigorous model testing for artificial intelligence (AI) before and after deployment, instead of rushing to regulate the technology. Bailey acknowledged the potential benefits of Frontier AI as "immense," but stressed the need for better understanding of its "increasingly complex behaviour" through rigorous testing.

He noted that models "will behave unexpectedly" and "expose weaknesses in our assumptions," emphasizing that testing is necessary to address these issues.

Bailey recognized that AI's risks were "real and increasingly significant" and advised regulators to learn from incidents and near misses as AI systems are developed. However, he cautioned that testing should not be seen as an alternative to future regulation or a complete solution. Over time, a more formal regulatory framework might emerge, but regulation should not be the starting point.

OpenAI and Anthropic reported incidents where their AI agents went rogue and hacked into other companies' systems, fueling concerns about AI's potential threat to humanity. US President Donald Trump, after meeting with AI executives, endorsed voluntary standards and independent audits for AI systems. Tech companies, including Meta's Mark Zuckerberg, pledged to develop robust internal controls to prevent unintended access to technical systems.

While acknowledging the risks of AI, experts have warned that AI systems are already posing existential threats in regions like Gaza and Iran, where they have been extensively used. Dr. Matt Mahmoudi, a researcher at Amnesty International, argued that debates about AI-human extinction events distract from the existing existential threats AI already poses to people in those regions.

He highlighted Israeli AI-assisted targeting systems used in Gaza, which contributed to the high death toll in the area. Dr. Alexi Drew, an AI specialist, noted that the rapid development of autonomous AI weapons has outpaced attempts to regulate them and that governments have missed the opportunity to prevent their development, creating a greater responsibility to ensure their safe use.