Three British intelligence agencies warn the government that artificial intelligence threatens the national security system

The security briefing received by the British Cabinet brought the leaders of the three intelligence agencies to a single table, delivering an explicit warning about the accelerating pace of AI models and their direct implications. Ken McCallum, the director-general of the Home Office, sat alongside Bliss Metroville, chief of the Secret Intelligence Service, and Ann Keyst Battler, director of the Government Communications Centre, to present a unified intelligence assessment that views these technologies as among the greatest threats to national security in the coming years, amid serious concerns that they could be exploited by hostile states, extremist groups and individuals seeking to damage digital infrastructure.

Artificial intelligence is the most important thing you should be thinking about today.With that direct statement, McCallum addressed members of the British government, according to The Times, prompting one of the present ministers to liken the current stage to the early phases of the Covid-19 pandemic and the global financial crisis of 2008, when warning signs emerged early without sufficient preventive measures. Although the briefing was part of an annual security-threat review rather than a response to an imminent danger, the sharp acceleration in model development generated concern among officials that existing policies are unable to keep pace with the rising risks.

Our report traced how the intelligence concern rested on documented field incidents rather than theoretical worries; the evidence included breaches carried out by autonomous AI models against corporate and institutional systems, cyber attacks targeting the banking sector in South Korea, and another case in which a model submitted a false murder report directly to a U.S. police website. Domestically, tests by the Government Cyber Security Coordination Centre over the summer uncovered 407 vulnerabilities across nine government bodies, including a critical flaw that allows an attacker to modify source code and control a major digital service.

This revelation of source-code-tampering vulnerabilities explains London’s intention to use next year’s G20 summit in Manchester to push for an agreement on a global code of conduct for AI use, aiming to reap its benefits and curb its harms.Shifting from manual exploitation of vulnerabilities to automated attacks reshapes the notion of digital sovereignty.The British diplomatic move demonstrates that advanced cyber-risk management requires coordination frameworks that transcend the traditional borders of each nation.

For decision-makers and technical teams in the Gulf, Egypt and the Arab region, this warning reshapes cyber-security priorities for financial and governmental institutions. The entry of autonomous models into the cyber-attack pipeline and their targeting of banks means that the cost of technical insurance will inevitably rise, and security teams are no longer concerned only with monitoring ports and networks; they must now continuously audit automation environments and source code to detect any automated manipulation. The threat goes beyond data leakage to the possibility of disabling or altering critical digital systems, necessitating advanced skills in independent code auditing to keep pace with the spread of AI agents in workplace environments.

Efforts to draft a global code at the upcoming Manchester summit place regional institutions before an imminent regulatory requirement: review your code standards and security policies for your models before international standards become sudden constraints on your cross-border digital activities. Proactive caution and code auditing are the real safety valves for protecting sensitive infrastructure from technologies that evolve faster than the regulations governing them.