UK foreign secretary warns delay on AI rules risks ‘Hiroshima’

UK Foreign Secretary Yvette Cooper warned governments risk an “AI equivalent of Hiroshima” if they delay international rules on artificial intelligence. In an article published Monday, she urged the United States, China and other major AI powers to agree shared safety standards.

Cooper wrote that AI offers major breakthroughs but also creates threats to security, public safety and social cohesion. She called the challenge “the greatest security challenge of the next decade” and compared the moment to the early nuclear era, when international agreements followed the bombing of Hiroshima. “We cannot afford to wait for an AI equivalent of Hiroshima before we act,” she wrote.

During a recent visit to Shenzhen, Cooper observed AI and robotics used in life-saving medical care. She noted those technologies are also being adapted in ways that affect warfare, crime and community trust.

Cooper asked Britain to use its diplomatic influence to convene talks between the US, China and other AI-leading countries to set shared safety principles and technical standards. She pointed to the UK’s 2023 AI Safety Summit at Bletchley Park, where leaders from 29 countries and the European Union met to discuss risks from advanced AI systems.

Several organisations have warned about concrete dangers tied to advanced models. The UK’s AI Security Institute reported rapid improvements in AI could enable automated cyberattacks after models completed simulated attacks without human assistance. The International Monetary Fund said AI could amplify cyberthreats to the financial system by lowering the skill needed to exploit weaknesses and urged treating cybersecurity as a financial stability issue.

In the US, President Donald Trump signed an executive order in June that set a voluntary framework for reviewing advanced AI models before release, expanded federal AI cybersecurity efforts and directed agencies to assess national security risks from frontier models. In industry, Anthropic’s chief executive Dario Amodei called for mandatory third-party testing of the most powerful systems.

The US government ordered Anthropic to restrict access to its Claude Fable 5 and Mythos 5 models on national security grounds before lifting that order in July. Regulators have cited tensions between rapid model development and national security concerns.

Cooper urged negotiators to agree on principles and testing regimes that could be adopted across jurisdictions so companies and governments operate to common minimum standards. Her appeal follows calls from governments, financial institutions and parts of the AI industry for coordinated oversight of frontier systems.