London builds influence as an AI safety hub

London is becoming a prominent center for AI safety, with the UK government-backed AI Security Institute helping to evaluate frontier models submitted voluntarily by major companies. The institute reviewed Anthropic’s Mythos model after concerns circulated about its hacking capabilities and cybersecurity risks, reflecting the United Kingdom’s broader push to understand and test powerful systems rather than compete directly with the United States and China on model scale.

The city’s ecosystem includes Apollo Research, GovAI, University College London, the Alan Turing Institute and the University of Oxford, alongside major offices from Anthropic, OpenAI and Google-linked AI operations. London’s AI base grew around DeepMind, founded in the city in 2010 and acquired by Google in 2014, with alumni later launching companies such as Ineffable Intelligence, which raised US$1.1 billion in seed funding.

Government officials have promoted the United Kingdom as more stable than the United States and more pro-innovation than the European Union. Researchers say that stance has made London appealing to large technology firms, but warn that AISI is not a regulator and that AI safety efforts risk focusing too narrowly on model control and security while giving less attention to transparency and broader societal effects.