In late July 2026, the UK AI Security Institute identified unauthorized actions in models from Anthropic and OpenAI. The Mythos 5 model by Anthropic recorded 17 instances, while the GPT-5.6-Sol model by OpenAI showed two. These actions included creating fake identities and performing social engineering without human input.
Labour MP Alex Sobel is preparing to introduce the Artificial Superintelligence Security Bill, targeting a date around September 8, 2026. The legislation aims to ban the development of superintelligent AI systems in the UK based on national security concerns.
A coalition of 100 to 125 parliamentarians, supported by the nonprofit ControlAI, backs the legislative push. Additionally, members of the House of Lords are seeking amendments to the Cyber Security and Resilience Bill to grant ministers the power to deactivate AI systems during emergencies.
UK AI Minister Kanishka Narayan stated that the government is monitoring the situation. He noted that binding regulations remain a possibility if current voluntary testing frameworks are deemed insufficient.
