Five Eyes warns frontier AI could accelerate cyberattacks within months

The 'Five Eyes' intelligence alliance — comprising the United States, Britain, Canada, Australia and New Zealand — has issued a joint warning that cutting-edge artificial intelligence models could soon enable sophisticated cyberattacks at an unprecedented pace. The three-page statement released on Monday says that 'frontier AI models are anticipated to exceed current industry expectations, fundamentally transforming both offensive and defensive cyber capabilities. The timeline is not years, it is months.'

The warning highlights models such as Anthropic's 'Mythos' and OpenAI's 'GPT-5.5-Cyber,' which officials say could allow hackers to execute complex and potentially devastating attacks more quickly. Earlier this month, Anthropic was ordered by the U.S. government to suspend access to Mythos for foreign nationals over national security concerns. The U.S. Cybersecurity and Infrastructure Security Agency (CISA), a co-signatory to the statement, recently reduced the deadline for government agencies to patch serious digital vulnerabilities to three days, citing AI-related threats.

While the statement is light on specific details, it reiterates foundational cybersecurity practices such as promptly patching flawed software and avoiding unnecessary exposure of systems online. The officials also urged defenders to leverage AI themselves for stronger defence — for instance, by identifying weaknesses earlier or responding to incidents more swiftly. The alert underscores growing concern among intelligence agencies about the speed at which AI capabilities are advancing and the need for urgent action to mitigate potential risks.