
The UK will be increasingly exposed to AI disruption unless it takes practical steps to strengthen its AI sovereignty, according to a new policy roadmap we’ve published today. Read more: bit.ly/48sp6QN #AI #AiSovereignty
✓The Alan Turing Institute is the UK's national institute for AI and data science.

The UK will be increasingly exposed to AI disruption unless it takes practical steps to strengthen its AI sovereignty, according to a new policy roadmap we’ve published today. Read more: bit.ly/48sp6QN #AI #AiSovereignty

“Threats are evolving faster than our ability to respond.” At our Securing Resilience for the Age of AI event, CEO George Williamson set out why the UK needs new AI capabilities to strengthen national resilience: bit.ly/4jmnQEZ

How can research & innovation help secure UK resilience for the age of AI? Our paper proposes a national agenda to turn cutting-edge research into practical capabilities that help us anticipate disruption, respond to threats & retain the freedom to act: bit.ly/4jkNAS6

We’re looking for two Distinguished Scientists to join our Fundamental Research team. Conduct internationally ambitious research & lead multidisciplinary teams: ➡️Sovereign Language Modelling: bit.ly/4k8lVox ➡️AI for Physical Systems: bit.ly/4k7GDVx

🚢 How do we know an autonomous ship is safe? Our new paper explores the assurance gap in the IMO's MASS Code. It argues for a shared framework to make autonomous vessel safety demonstrable, consistent and auditable. ➡️bit.ly/4ApMQkv @LR_Foundation @lloydsregister

How can open-source AI help government make sense of complex real-world data? Open Source AI Fellow, Armin Mustafa, shares how she's working with government departments to develop AI tools suitable for high-stakes settings like the police force: bit.ly/3Vhxek4

Join an exciting, new Defence AI engineering team this #NationalEngineeringDay Join us to: ➡️Tackle complex, disparate data ➡️Build pipelines & tools ➡️Help analysts & decision-makers act with confidence Discover where you can excel – apply now: bit.ly/4sp376p

We’re hiring: help turn cutting-edge research into capability for defence & national security Join us to: ➡️Tackle complex, disparate data ➡️Build pipelines & tools ➡️Help analysts & decision-makers act with confidence Apply: bit.ly/4sp376p

New model presented today at the European Radar Conference 2026: enhancing autonomous space domain awareness Explore how our Defence AI Research Centre (DARe) & @unibirmingham's new model improves identification of objects in orbit: bit.ly/4BfylBd #WSW2026

In his blog, co-author of the paper Rick Hennessy explores the obedience paradox, alongside some of the agentic AI behavioural failure modes that were seen in the Hugging Face hack earlier this year: bit.ly/4z0xNgx

How to assure agentic AI behaviour in high-stakes settings: beyond the Hugging Face incident A new briefing paper from the Turing’s Centre for Emerging Technology and Security (CETaS) examines how we can assure the behaviour of agentic AI: bit.ly/3V7ecws

CETaS has also published a new briefing on behavioural assurance for agentic AI. The paper examines how AI agents can behave contrary to organisational intentions & explores how organisations can better understand, monitor and govern agent behaviour. 📄bit.ly/3V7ecws

Today we're also launching a new £2m research programme on transformative AI security risks, delivered by our Centre for Emerging Technology & Security. Funded by @coeff_giving, it will examine how transformative AI could reshape the security landscape. 🔍bit.ly/4As4S6l

The report argues that stronger AI verification & assurance will be essential if we are to understand whether AI-enabled systems remain dependable once deployed in real-world settings. As AI systems become more capable, we need better ways to assess and manage risks in practice.

Frontier AI Risks: A practical way forward highlights five frontier AI risks where stronger action can be taken now: 🔹Cyber security risks 🔹Democratic instability 🔹Misalignment 🔹Loss of meaningful human control 🔹Chemical and biological threats 📖bit.ly/4hqXHmr

What should we do about frontier AI risks when there is still uncertainty about the most extreme long-term scenarios? Our new report argues that uncertainty should not delay practical action. Many AI capabilities, behaviours & harms can already be observed, studied & addressed.