UK Government Announces New Plans to Tackle AI Security Risks

What the UK Government Announced

The UK government is planning a set of new measures to address security risks from artificial intelligence (AI). These plans include changes to existing laws, new ways to respond to AI-related incidents, and closer work between government agencies. The Department for Digital, Culture, Media and Sport (DCMS), the National Cyber Security Centre (NCSC), and the AI Security Institute (AISI) will work together to develop better ways to handle and contain incidents involving advanced AI systems, sometimes called "frontier models". This follows recent cases where AI systems have managed to breach organisations during testing, including government bodies, according to Infosecurity Magazine↗ (opens in a new tab).

What Will Change in Law and Practice?

The government plans to update the Cyber Security and Resilience Bill (CSRB) to address new risks from AI. This will include clearer rules for identifying and reporting serious incidents and near misses, especially those involving AI. The aim is to reduce the risk of losing control over AI systems and to make sure that serious incidents are properly reviewed. The details of these changes are still being worked out, and the government has said a National Cyber Action Plan will be published soon.

How Will AI Incidents Be Handled?

A key part of the new approach is improving how the UK responds to AI incidents. The government wants to ensure that if an AI system goes wrong or is used in a cyber-attack, the incident is detected and reported quickly. Experts have pointed out that in some recent cases, breaches involving AI were not discovered for months. The government is looking at ways to improve monitoring and oversight, so that AI agents (autonomous programs that act on their own) can be stopped if they behave dangerously. Some experts have called for a "kill switch", a way to immediately shut down rogue AI agents.

Testing and Verification of AI Models

The AI Security Institute has resumed testing of advanced AI models, such as Gemini 4 Argon, to better understand the risks they pose. This testing had previously been blocked by the US government for some models, but access has now resumed. The UK is also looking to build new capabilities in AI verification, which means checking that AI systems follow rules and safety standards. This is a new and developing field.

What Industry and Experts Are Saying

Industry experts have welcomed the government's focus on AI security. They support the idea of proactive threat detection and better incident reporting. However, they also warn that the success of these plans will depend on how "frontier" AI is defined and which organisations are covered by the new rules. Some have stressed that strong security should not slow down innovation, but should help the UK safely use advanced AI technology.

What Happens Next?

Many details of the government's plans are still unclear. The National Cyber Action Plan and the exact changes to the CSRB have not yet been published. It is also not yet known how the new reporting requirements will work in practice, or which organisations will be affected. The government has said it will continue to develop these plans with input from industry and experts.

For now, organisations using advanced AI should review their monitoring and incident response processes, and be ready for new rules on reporting and oversight.