**London Emerges as a Key Hub for AI Safety and Governance**
When Mythos, an artificial-intelligence model from US firm Anthropic, launched to a limited group of companies earlier this year, rumours quickly circulated about the system’s hacking capabilities and threats to cybersecurity. Real clarity about the model’s power came only once the AI Security Institute (AISI) — a UK government-backed organization in London — had evaluated the model.
London, particularly its technology quarter in King’s Cross, is emerging as a key hub outside the United States for AI firms (see ‘King’s Cross hub’), with Anthropic and OpenAI, both headquartered in San Francisco, California, opening major offices there. And the city’s approach has a distinct flavour to that of its international counterparts. Rather than competing with the United States and China on generating ever bigger and more-powerful models, the United Kingdom is carving out a focus on AI safety and understanding the risks of emerging models. AISI — which vets frontier models that companies submit to it on a voluntary basis — is one of several UK AI safety laboratories operating globally.
“The UK has been out ahead in thinking about ethics, responsible AI and AI governance for quite a while,” says David Leslie, who researches ethics, technology and society at Queen Mary University of London.
The London-based safety ecosystem also includes Apollo Research, a lab that has carried out high-profile research into the potential for models to ‘scheme’ against human users, and non-profit organization GovAI, formerly the Centre for the Governance of AI, which has co-authored influential research, such as proposing using access to computing resources as a way to govern AI.
Academic bodies such as University College London, the Alan Turing Institute and the University of Oxford are also nearby, providing AI research expertise and graduate talent.
Although a focus on AI safety is proving to be a successful niche for the United Kingdom, some researchers highlight that AISI is not a substitute for a regulator. Others are also concerned that a narrowing definition of AI safety has become too focused on keeping models under control and secure from attacks, while paying less attention to wider societal impacts such as transparency and the consequences of widespread adoption.
**London calling**
London’s broader AI ecosystem has been built around DeepMind, the AI firm founded in the city in 2010 and acquired by Google in 2014. Its co-founder, Demis Hassabis, ensured that the firm stayed in London despite the sale. Several of its alumni have founded major start-ups in the city, including Ineffable Intelligence, established by AI researcher David Silver, which earlier this year raised US$1.1 billion in seed funding.
Being in the United Kingdom “gives us a bit more space to think through things in a slightly different way than if you’re in the wheelhouse the whole time”, said Hassabis, speaking to *Nature* in 2025. Although “hype culture” is growing in the UK AI community, “there’s always been a more measured and an evidence-based approach to evaluating science”, adds Leslie.
The UK government is trying to entice foreign companies by pitching the country as more stable than the United States and more pro-innovation than the European Union. In March, after Anthropic became embroiled in a spat with the US Department of Defense over ethics, London’s mayor Sadiq Khan wrote to the company applauding its stance and inviting it to expand in London, which the firm did the following month. Tech firms see the United Kingdom as being more “amenable to their interests than the EU” is, which has the relatively strict EU AI Act, says Leslie.
“The big tech companies feel as though they have not only a rich environment to build from in terms of intellectual talent, but they’ve got a very agreeable regulatory environment and a very agreeable government environment” in London, he says.
**Safety first**
When it comes to efforts to chart the risks of AI, London’s standout resident is the AISI. It was launched in 2023 by former prime minister Rishi Sunak at the first global AI Safety Summit, held at Bletchley Park in Milton Keynes, UK, when the country was at the forefront of an international push for AI safety.
It came out of a “really big belief in governments needing to understand what is going on” at the AI frontier, says Jade Leung, the organization’s chief technology officer. AISI performs industry-wide evaluations, such as how reliably models act as intended or are able to persuade the public. Firms such as Google DeepMind, OpenAI and Microsoft also submit their models for testing, for example to check their capabilities in creating biological weapons. AISI researchers try breaking down the model’s safeguards, so that the firm can improve them.
—
## FAQ
**Q1: What is the AI Security Institute (AISI) and what does it do?**
AISI is a UK government-backed organization that evaluates the safety of artificial-intelligence models. It tests frontier models for reliability, alignment, and potential risks, including misuse capabilities such as hacking or persuasion. Companies submit their models voluntarily for assessment, and AISI researchers attempt to break model safeguards to help firms improve security.
**Q2: Why is London becoming a major hub for AI firms?**
London offers a combination of top-tier AI research institutions, a strong talent pool, a supportive regulatory environment, and proximity to key markets. The UK government has positioned the country as more stable than the US and more innovation-friendly than the EU, attracting firms like Anthropic and OpenAI to establish major offices in King’s Cross.
**Q3: How does the UK differ from the US and China in AI development?**
While the US and China often focus on building larger, more powerful models, the UK has carved out a niche in AI safety and governance. Organizations like AISI prioritize understanding and mitigating risks, emphasizing ethics, transparency, and responsible deployment over sheer model scale.
**Q4: What concerns exist regarding AI safety in the UK?**
Some researchers argue that AISI is not a substitute for formal regulation. Others worry that an overly narrow focus on model security and control may overlook broader societal impacts, such as transparency, accountability, and the consequences of widespread AI adoption.
**Q5: Which notable AI organizations are based in London?**
London is home to DeepMind (founded there, later acquired by Google), as well as alumni-founded companies like Ineffable Intelligence. The city also hosts research labs such as Apollo Research and organizations like GovAI, along with strong academic institutions including University College London, the Alan Turing Institute, and the University of Oxford.
—
## Conclusion
London is rapidly establishing itself as a global center for AI safety and governance, offering a distinctive alternative to the AI development models seen in the United States and China. With a robust ecosystem of research institutions, safety-focused organizations like the AI Security Institute, and a regulatory environment that balances innovation with oversight, the UK is positioning itself as a leader in responsible AI. However, ongoing debates about the scope of AI safety and the need for stronger regulatory frameworks highlight that the journey is only just beginning. As the AI landscape evolves, London’s emphasis on ethics, transparency, and risk management may well define the next chapter in artificial intelligence development.
