Spy agencies warn AI could supercharge hacking within months, not years

The world's leading intelligence-sharing alliance has issued an unusually blunt warning: advanced artificial intelligence is about to make hacking far more dangerous, and the window to prepare is shrinking fast. In a joint statement released on Monday, cybersecurity agencies from the United States, United Kingdom, Canada, Australia, and New Zealand, collectively known as the Five Eyes, said that so-called "frontier" AI models are set to reshape both offensive and defensive cyber capabilities faster than most organizations expect. The agencies were direct about the timeline, stating plainly that the shift is coming in months, not years.

The three-page statement, while attention-grabbing in tone, was fairly conservative on specifics. Much of the practical advice it offered was familiar ground for cybersecurity professionals: patch software quickly, avoid putting unnecessary systems online, and tighten up basic digital hygiene. The agencies also pushed organizations to flip the script and use AI defensively, suggesting it could help spot vulnerabilities sooner and speed up responses when attacks do happen.

Also Read: Mobile phone services suspended for Muharram security

The timing of the warning is hard to separate from the recent controversy surrounding specific AI systems. Officials pointed to models like Anthropic's Mythos and OpenAI's GPT-5.5-Cyber as examples of tools capable of executing complex, potentially devastating hacking operations with far less effort than was previously required.

That concern has already translated into concrete action. Earlier this month, Anthropic was ordered by the US government to cut off foreign nationals' access to certain versions of Mythos, citing national security concerns. The company complied, disabling the affected access, though it's a move that underscores how seriously Washington is treating the risk of these tools falling into the wrong hands. Around the same period, the US Cybersecurity and Infrastructure Security Agency, known as CISA and one of the signatories to Monday's statement, slashed the deadline given to federal agencies for fixing serious software vulnerabilities down to just three days.

The most striking line in the statement was a warning that AI systems could soon begin improving themselves faster than humans are able to monitor or control, a scenario that has long been a hypothetical concern among AI researchers but is now being treated as a near-term planning consideration by national security officials.