Intelligence agencies from five countries - the U.S., United Kingdom, Canada, Australia, and New Zealand - have issued a stark warning that cybersecurity threats posed by artificial intelligence (AI) could become a reality within months, urging immediate action.

On Monday, the Five Eyes intelligence alliance released a joint statement, asserting that frontier AI models are poised to fundamentally transform both offensive and defensive cyber capabilities, surpassing current industry expectations. These changes are not years away, but could occur within mere months.

The alliance emphasized that world leaders must act without delay.

Specifically, they called for: understanding and assessing risks, preparedness, and accountability; prioritizing basic cybersecurity practices and controls; empowering cyber leaders with authority and resources; and maintaining active engagement as threats and guidance evolve.

The Five Eyes agencies stressed that a comprehensive response from organizations and society is crucial, arguing that cyber risks are no longer just technical issues but core business risks and leadership responsibilities.

While the statement didn't provide specific details about AI-related cybersecurity threats, it builds on the fundamental discussions that emerged after the Mythos Shock, which suggested AI could disrupt existing security paradigms.

However, the fact that intelligence agencies have issued direct warnings about AI-related cybersecurity threats lends significant weight to the issue.

<em>Reuters </em>noted that this warning is another sign of growing concerns among intelligence authorities regarding AI models such as Anthropic's Mythos and OpenAI's GPT-5.5-Cyber.

Although the statement didn't mention specific AI models, it is widely interpreted as targeting Anthropic's Mythos, which has recently been subject to export control measures.

On June 12, Anthropic reported that the U.S. government had issued export control guidelines to completely halt foreign access to Claude Mythos 5 and Claude Fable 5. While the South Korean government and some companies recently joined Project Glasswing, which provides access to Mythos, this U.S. administration action has effectively stalled the model's utilization.

Intelligence agencies from five countries - the U.S., United Kingdom, Canada, Australia, a …

AI Cybersecurity Threats: What the Five Eyes Alliance Warns About in 2026