Five Eyes intelligence alliance warns of threats from new types of AI | Cybersecurity

Frontier AI models are ‘significantly changing’ cyber security capabilities, intelligence officials say.

Technology is poised to increase the ability to hack people, and urgent action is needed to combat the threat, officials in the US, UK, Canada, Australia and New Zealand have said.

“Future AI demonstrations are expected to exceed the industry’s current expectations, and transform both offensive and defensive cyber capabilities,” the intelligence agency known as Five Eyes said in a three-page statement on Monday.

“It’s not years, it’s months.”

The text was light on details and reiterated Internet safety advice, such as uninstalling malicious software and keeping computers off the Internet unless absolutely necessary.

The officials also encouraged the opposition to use AI to “strengthen security”, for example by identifying weaknesses quickly or responding quickly to situations.

The warning was another sign of concerns that officials have over brands such as Anthropic’s Mythos or OpenAI’s GPT-5.5-Cyber, which they say allow users to quickly carry out complex — and destructive — hacks.

Earlier this month, Anthropic was forced to cancel the Mythos brand after the US government ordered it to freeze access to foreign samples due to national security concerns.

At the same time, the US Cyber-defense Agency, CISA, which was among the signatories of Monday’s statement, reduced the time given to government officials to deal with digital vulnerabilities in their networks to three days, citing AI threats.