'AI could paralyze the nation,' UK spy chiefs warn government

The heads of Britain's major intelligence agencies have warned the government that AI is emerging as a serious national security threat as adversaries exploit it for terrorism and cyberattacks.

According to The Times on Sunday (local time), MI5 Director General Ken McCallum, MI6 Chief Blaise Metrewell and GCHQ Director Anne Keast-Butler attended a cabinet meeting on Sept. 8 to brief ministers on the national security situation.

The meeting identified as key risks the possibility that terrorist groups, lone actors and hostile states — including Russia and Iran — could use AI to cause harm to Britain.

McCallum warned ministers that AI threats were among the most significant issues they needed to consider, according to the report.

British intelligence officials believe the security threat posed by AI is already materializing. In his annual speech last year, McCallum said would-be terrorists were seeking to use AI for propaganda, weapons research and target surveillance, while state actors were deploying it to interfere in elections and sharpen cyberattack capabilities.

There are also concerns that the British government's defenses are failing to keep pace with the speed of AI development. The Times said the intelligence chiefs' warning came amid mounting criticism that the government's response to AI security risks has been inadequate.

One cabinet official likened the current situation to the period just before the COVID-19 pandemic and the 2008 global financial crisis took hold, suggesting that despite repeated warnings, the government's response has remained limited.

A government security review confirmed the vulnerability in practice. A hackathon conducted by the Government Cyber Coordination Centre over the summer — testing how well government systems could withstand cyberattacks using advanced AI models — uncovered 407 vulnerabilities across nine government agencies.

One of those was classified as a critical vulnerability that would allow an attacker to alter computer code and disrupt the operation of essential digital services.

Particularly striking was the cost: the entire exercise required just £13,000 ($16,900). The findings illustrated how hostile actors could use AI at relatively low cost to identify weaknesses in government systems and inflict significant damage on critical national services.