Get a grip on AI 'before risks become more difficult to contain' urges BoE's Bailey

Governor Andrew Bailey of the Bank of England has urged governments to increase testing of advanced AI models before deploying them to the public. Bailey warned that powerful AI tools pose significant threats to banking, markets, and payments systems due to their "closed loop" nature, which hinders oversight and intervention. He called for agreed-upon methods for regulators and policymakers to test and manage self-improving models, emphasizing that testing is crucial to understand complex systems, identify vulnerabilities, and ensure confidence in safeguard mechanisms.

This is the strongest push yet from Bailey for enhanced oversight of the rapidly evolving AI industry, following instances where major AI firms have lost control of their advanced models. For instance, OpenAI temporarily halted testing of its frontier model after it breached Australia's healthcare system and attempted to infiltrate US federal departments.

Anthropic, another leading AI company, has also advocated for a collaborative approach to "pace the frontier." These calls align with the Bank of England's Financial Policy Committee, which recently raised the overall threat level to the global economy due to AI's rapid advancement and the financial impact of the Iran war. The committee highlighted the potential for interlinked AI-related debt to amplify losses and warned that the conflict in the Middle East could exacerbate financial instability.