LONDON / RankWire.AI / – Britain is taking steps to strengthen its national AI security framework after government-led testing uncovered over 1,300 software vulnerabilities. The assessments involved analyzing hundreds of code repositories utilized across various central government departments. Officials indicated that the use of advanced AI tools was instrumental in detecting cyber weaknesses and assessing potential threats. As a result, the government aims to expand this initiative across all public systems, including preparations for serious incidents involving autonomous AI technologies.
The two pilot programs uncovered and addressed a total of 1,352 vulnerabilities within 548 code repositories spanning 18 government departments. These teams employed cutting-edge AI systems to scan software and pinpoint areas of weakness that required urgent attention. Moving forward, departments will prioritize fixing flaws in their most critical and high-risk systems. Additionally, Britain seeks to increase its capacity to leverage frontier AI for cyber defence purposes. Officials stated that this initiative will bolster the security of government technology and digital services, making them more resilient against cyber threats.
Separately, the AI Security Institute is conducting evaluations of advanced AI models, focusing on whether their safety controls function as intended. Artificial Intelligence Minister Kanishka Narayan emphasized that over 100 technical specialists are currently working at the institute. Present assessments include Gemini 4 Argon and other sophisticated AI systems, with the testing aimed at understanding model capabilities, security risks, and the robustness of safeguards. Officials have reiterated that independent evaluations remain a vital component of the nation’s AI security strategy.
Expanding Cyber Response Framework
In addition to vulnerability testing, authorities are developing protocols for managing major incidents involving agentic AI systems, which are capable of executing complex tasks with minimal human oversight. The National Cyber Security Centre is collaborating with other government agencies to establish a comprehensive response framework, considering how authorities should detect, contain, and assist organizations impacted by such events. This work also involves assessing whether current cyber response systems can effectively handle failures involving increasingly autonomous AI technologies.
Furthermore, Britain is considering whether the Cyber Security and Resilience Bill should incorporate stronger measures to address AI-related threats. The scope under review includes serious incidents, near misses, and situations involving loss of control. Existing proposals already require certain regulated operators and digital service providers to report incidents, with initial notifications expected within 24 hours and comprehensive reports within 72 hours under the proposed framework.
Prioritizing AI Safety Measures in National Security
Narayan highlighted that Britain needs to bolster its cyber defenses by enhancing independent technical testing and cultivating domestic expertise in advanced AI. He also suggested legislation may be necessary if current safeguards prove insufficient for emerging risks. The government is working on additional support mechanisms for AI verification, with plans to publish more details in the upcoming months. Such verification techniques are crucial for ensuring systems operate correctly and that security controls remain effective, even under challenging or atypical conditions.
These initiatives form part of a broader government review of artificial intelligence and its potential impact on national security and societal infrastructure. The Prime Minister’s AI Taskforce is examining issues related to security, public service delivery, economic development, and workforce implications tied to AI technology. Moreover, Britain has scheduled AI to feature prominently during its 2027 G20 presidency. The latest measures are focused on remedying software vulnerabilities, assessing advanced AI models, and enhancing readiness to respond to significant AI-related incidents within government systems.