Skip to main content

London builds an AI-safety ecosystem around research and policy

22 NOVEMBER 2026·2 MIN READ·4 SOURCES

London is home to Google DeepMind and a major Anthropic base, while the London AI Hub estimates that up to 50 organisations in the capital focus on AI safety. The activity spans model evaluation, research, government investment and plans for a larger hub.

Key takeaways · 4

  • 01

    For model-risk planning, note that a state-backed organisation assesses leading models before and after release.

  • 02

    LISA’s planned headquarters is 82,000 square feet, with the move expected during 2027 after a fit-out.

  • 03

    The government’s additional £120 million for the AI Security Institute is intended to strengthen testing of advanced systems and understanding of their risks.

  • 04

    AI-biology work has potential benefits in medicine and vaccines, alongside concerns that it could make harmful pathogens easier to develop.

Research and model evaluation

The London AI Hub estimates that up to 50 organisations in the capital focus on AI safety.[1] A state-backed organisation assesses leading models before and after release for risks including cyber misuse.[1] Apollo Research examines whether models could act against human interests by lying or cheating.[1] One obstacle for researchers is access to cutting-edge models before public release.[1]

Government and AI-biology work

The UK government has announced hundreds of millions of pounds in new investment for AI safety and security, including work on cyber threats and the risk of advanced systems acting beyond human control.[1] An additional £120 million for the AI Security Institute is intended to strengthen the UK’s ability to test advanced systems and understand their risks.[1] The government also plans to launch an AI Biosecurity Lab.[1] Separately, Basecamp Research will work with OpenAI and Anthropic on antibiotics and vaccines intended to protect against future pandemics and help anticipate emerging threats.[1] Experts note that AI could accelerate medicine and vaccine discovery, but could also make harmful pathogens easier to develop.[1]

Capacity, constraints and debate

The ecosystem faces practical constraints alongside questions about how quickly AI should advance. Building a world-leading AI ecosystem will require data centres, and energy use is a concern; energy is described as the single most critical constraint on London’s data-centre growth.[1] Brady estimates that more than 750 people at Imperial work directly on large language models or their applications.[1] Pause AI UK calls for a halt to frontier AI development until models can be made safe, and the group says hundreds of people are joining every month.[1] Prof Hannah Fry said she remained optimistic, describing London as focused on ethical technology and “things people care about.”[1]

London has a state-backed organisation assessing leading models, the government plans to launch an AI Biosecurity Lab, and Basecamp Research will work with OpenAI and Anthropic on antibiotics and vaccines. Researchers face challenges gaining access to cutting-edge AI models before public release, and building a world-leading AI ecosystem will require data centres; energy is the single most critical constraint on London’s data-centre growth. Pause AI UK calls for a halt to frontier AI development until models can be made safe.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 22 November 2026

    London builds an AI-safety ecosystem around research and policy

Sources

AI fluency, one session a day, built for your work.