LONDON / RankWire.AI / – Britain is broadening its defenses against the threats posed by advanced AI following tests that uncovered over 1,300 cyber vulnerabilities. These weaknesses were identified across hundreds of government software repositories. The UK government stated that its initiative will leverage cutting-edge AI tools to enhance cybersecurity and pinpoint vulnerabilities within vital digital infrastructure. Preparations are also underway for responding to serious incidents involving autonomous AI systems. This effort is part of a larger strategy to bolster national resilience as more sophisticated AI models become accessible to the public.
In two government-led pilots, a total of 1,352 vulnerabilities were discovered and mitigated across 548 code repositories in 18 departments. These trials aimed to evaluate how advanced AI could assist in detecting software flaws and recognizing potential cyber threats. Moving forward, officials intend to expand the frontier AI hardening initiative across various government sectors, focusing on systems with the highest levels of risk. The programme also seeks to develop broader capabilities for AI-driven cyber defence within the government while ensuring proper oversight of the emerging technology.
The Artificial Intelligence Minister Kanishka Narayan highlighted that Britain’s AI Security Institute is actively assessing advanced models and testing the robustness of their safeguards. The institute, which employs over 100 technical specialists, is currently evaluating models such as Gemini 4 Argon and other high-level systems. The government emphasizes that independent testing remains essential because public authorities cannot solely depend on AI developers to evaluate the risks associated with increasingly capable models.
Expansion of AI incident response planning
Collaboration between the National Cyber Security Centre and the AI Security Institute is advancing protocols for major incidents involving agentic AI. These autonomous systems are capable of executing complex tasks with minimal human oversight. The upcoming framework aims to guide authorities through incident detection, risk containment, and support for affected parties. Officials are also reviewing how current cyber response structures can address failures involving autonomous AI and other advanced technologies critical to essential services.
Additionally, the government is contemplating amendments to the Cyber Security and Resilience Bill progressing through Parliament. Proposed measures include enhanced treatment of AI-related cyber risks, serious incidents, near misses, and events involving loss of control. Existing proposals mandate that regulated operators and digital service providers report significant cyber incidents—initial notifications are to be submitted within 24 hours, with detailed reports due within 72 hours. The security framework encompasses vital services such as hospitals, energy grids, and drinking-water facilities.
UK reinforces AI safety protocols
Narayan stressed that the UK must combine fortified cybersecurity with independent testing and bolster domestic technical expertise. He also mentioned that legislation might be necessary where current safeguards are insufficient against emerging risks. The government has initiated efforts to support AI verification and plans to release further details in the coming months. Such verification tools are crucial for confirming that advanced systems behave as intended and that safety measures remain effective under demanding conditions.
These measures are part of a broader government initiative to assess the national security and public service implications of artificial intelligence. The Prime Minister’s AI Taskforce is actively reviewing issues related to economic growth, public services, workforce impacts, and security concerns. Britain has also designated AI as a key policy focus for its 2027 G20 presidency. The latest cybersecurity actions are aimed at testing advanced AI models, fixing software vulnerabilities, and preparing government systems to handle serious AI-related incidents.