UK to work with other 'middle powers' on AI security, minister says

Britain will work ​with other "middle powers" such ​as France, Germany and Canada ​on AI security to help shape the future of a technology dominated by the United ‌States and ⁠China, ⁠Technology Secretary Liz Kendall said.

Britain's AI Security ​Institute (AISI) generated interest from other countries when it evaluated ​Anthropic's Claude Mythos model earlier this month, showing that it could exploit ​security weaknesses, she said on ⁠Tuesday.

Kendall ‌told reporters: "Many other ​countries are extremely ​interested in working with ⁠us to learn some of the lessons ​from what AISI is doing."


Britain will publish best practice on evaluating AI models in July at the next meeting of AI Security Institutes.

AISI, which sits ‌under Kendall's department, has agreements with OpenAI, Anthropic and others to ​evaluate ​their models.

Kendell ⁠said work with Britain's peers should not be seen as weakening its "deep, close and enduring relationship" with the United States, which is home to the five hyperscalers that control about 70% of global AI compute power.