LONDON / RankWire.AI / – Britain is expanding national defences against advanced AI risks after trials found more than 1,300 cyber vulnerabilities. The weaknesses appeared across hundreds of government software repositories. The British government said its programme will use frontier AI tools to strengthen cyber security and identify weaknesses in critical digital infrastructure. Officials are also preparing measures for serious incidents involving autonomous AI systems. The work forms part of a wider effort to strengthen national resilience as more capable AI models enter public use.
Two government pilots identified and addressed 1,352 vulnerabilities across 548 code repositories in 18 departments. The trials tested how advanced AI systems could help locate software weaknesses and detect potential cyber threats. Officials now plan to extend the frontier AI hardening programme across government. Departments will focus on weaknesses in systems carrying the highest risk. The programme also aims to build wider government capabilities for using AI in cyber defence while maintaining oversight of the technology.
Britain’s AI Security Institute is evaluating advanced models and testing the effectiveness of their safeguards. Artificial Intelligence Minister Kanishka Narayan said the institute has more than 100 technical specialists working on model evaluation and security. Its current work includes assessments of Gemini 4 Argon and other advanced systems. The government says independent testing remains necessary because public authorities cannot rely only on AI developers to assess risks from increasingly capable models.
AI incident planning expands
The National Cyber Security Centre is working with government departments and the AI Security Institute on procedures for major incidents involving agentic AI. These systems can carry out complex tasks with limited human supervision. The planned framework will guide authorities as they identify incidents, contain risks and support affected parties. Officials are also examining how existing cyber response structures can address failures involving autonomous AI systems and advanced technologies across essential services.
The government is also considering changes to the Cyber Security and Resilience Bill as it progresses through Parliament. Potential measures include stronger treatment of AI-related cyber risks, serious incidents, near misses and loss-of-control events. Existing proposals require certain regulated operators and digital service providers to report major cyber incidents. Operators would submit initial notifications within 24 hours and fuller reports within 72 hours. The wider security framework covers essential services including hospitals, energy networks and drinking-water systems.
Britain strengthens AI safeguards
Narayan said Britain must combine stronger cyber defences with independent testing and domestic technical capability. He also said legislation may be needed where existing safeguards do not adequately address new risks. The government has begun developing support for AI verification and plans to provide further details in coming months. Officials say verification tools can help determine whether advanced systems behave as intended. They can also assess whether safety controls continue working under demanding conditions.
The measures sit within a wider government programme examining the national security and public-service implications of artificial intelligence. The Prime Minister’s AI Taskforce is reviewing issues linked to economic growth, public services, the workforce and security. Britain has also identified AI as a major policy issue for its 2027 G20 presidency. The latest cyber security measures focus on testing advanced models, repairing software vulnerabilities and preparing government systems for serious AI-related incidents.