Britain’s intelligence chiefs have warned that artificial intelligence (AI) could become one of the biggest threats to UK national security, with hostile states, terrorist groups and lone actors potentially using increasingly advanced AI to launch cyberattacks and cause serious damage.
Sir Ken McCallum, the MI5 director general, reportedly told ministers that AI was “the biggest thing you should be thinking about”.
McCallum, MI6 chief Blaise Metreweli and GCHQ director Anne Keast-Butler briefed the cabinet on Sept. 8 as part of a regular assessment of the national threat picture. The meeting was not prompted by an imminent danger, but included an extended discussion about how AI could be exploited by hostile actors, including Russia and Iran, as well as terrorist organizations and individuals.
According to a report by The Sunday Times, concerns among senior government figures have focused on the rapid development of AI models and whether the government is doing enough to address the risks.
One cabinet minister reportedly compared the situation with the early stages of the COVID pandemic and the 2008 financial crisis, when warning signs were visible but limited action followed.
McCallum had previously warned that terrorists were already seeking to use AI for “their propaganda, their weapons research, their target reconnaissance”, while state actors were using AI to manipulate elections and strengthen cyberattacks.
He also acknowledged that AI presented opportunities for British intelligence agencies, but cautioned that “it would be reckless to ignore the potential for it to cause harm”.
The warnings come amid a series of incidents highlighting how AI capabilities are being used in cyber and other harmful activities.
407 Vulnerabilities Found in UK Govt Systems
OpenAI disclosed this year that its AI agents had autonomously hacked the computational tools company Hugging Face. OpenAI later acknowledged that an AI agent had hacked into Australia’s Medicare system, prompting Prime Minister Anthony Albanese to describe the delay in alerting Australian leaders as “unacceptable”.
Banks in South Korea also suffered a series of cyberattacks that appeared to have been launched by a hacker using Chinese and American AI tools. Separately, an Anthropic AI model submitted a false homicide tip to a Philadelphia police website, reportedly marking the first known case of a bogus AI tip being sent to authorities.
Britain has already identified vulnerabilities that could potentially be exploited by advanced AI. During a series of government “hackathons” this summer, the Government Cyber Co-ordination Centre uncovered 407 security vulnerabilities across nine government organisations.
One vulnerability was classified as “critical” because it could allow an attacker to alter computer code and affect the operation of a “key digital service”. The service was not identified publicly. The exercises cost just £13,000, highlighting the relatively low cost at which sophisticated attacks could potentially be carried out.
Prime Minister Andy Burnham has said Britain will put AI “centre stage” during its G20 presidency next year and seek to develop a “global code” for AI use. The government has also established a £500 million Sovereign AI Fund over five years to invest in British AI companies, including those with cybersecurity capabilities.
A government spokesman said: “National security is the first duty of government.” The spokesman added that as AI becomes more powerful, Britain must be prepared for the threats it brings by testing advanced systems, building domestic capabilities and “hardening our defences”.
Leave a comment