Britain has urged a joint international response, warning that delaying regulation of artificial intelligence could repeat policy failures seen in the early days of nuclear weapons development.

According to blockchain outlet Decrypt on July 6, British Foreign Secretary Yvette Cooper (이베트 쿠퍼) said an international agreement to manage AI risks must be put in place quickly.

In a statement released that day, Cooper noted that AI is driving innovation in healthcare and robotics, but that rapid advances and wider adoption are also creating new risks for war, crime and social integration.

She said she recently looked firsthand in Shenzhen, China at AI and robotics technologies used in medical settings and assessed that she had confirmed "astonishing potential." She warned, however, that "the same technologies are changing the future of war and crime, and of social cohesion, in worrying ways."

Cooper stressed that managing AI risks could become "the biggest security challenge" over the next 10 years. She argued it is too late to draw up regulation only after problems emerge and that international safety principles and shared standards for advanced AI should be established first.

She compared today’s AI development race to the early nuclear arms race. Noting that a full-scale international agreement on nuclear technology came only after the atomic bombing of Hiroshima, she said, "We cannot afford to wait for an AI Hiroshima."

She also proposed that Britain should use its diplomatic influence to bring together major AI developers including the United States and China and help establish shared safety standards.

Cooper cited the AI Safety Summit held in 2023 at Bletchley Park in Britain as a leading example. Leaders from 29 countries and the European Union gathered to discuss emerging AI risks, which she assessed as "an example showing Britain can bring the world together for AI security."

Her remarks also come as concern grows over safety as AI systems rapidly become more advanced. In May, the UK AI Security Institute said OpenAI’s GPT-5.5 became the second AI model to carry out a simulated cyberattack without human help, warning that AI cyberattack capabilities are improving quickly. The earlier model with the same level of capability was Anthropic’s Claude Mythos Preview.

The International Monetary Fund also warned that AI could lower the technical barriers needed for cyberattacks and amplify attacks targeting the global financial system, calling for cybersecurity to be treated as an issue of financial stability.

The United States has also moved to respond. President Donald Trump signed an executive order last month to set up a voluntary framework for reviewing advanced AI models before release and to expand AI cybersecurity programs. It also requires government agencies to evaluate advanced AI models that could affect national security.

In the AI industry, calls for regulation have continued. Anthropic CEO Dario Amodei (다리오 아모데이) has argued that mandatory third-party safety testing is needed for advanced AI models. The U.S. government later ordered Anthropic to restrict access to Claude Fable 5 and Mythos 5, citing national security concerns, and the measures were lifted in July.

As competition in AI technology expands into healthcare, finance and national security, Britain has made clear that establishing safety standards in advance through international cooperation should take priority over after-the-fact responses. An analysis said the key is whether major countries can avoid seeing AI only as an object of industrial competition and instead build safety and security frameworks at a global level.