Five Eyes agencies say frontier AI models near advanced cyber capabilities

Cybersecurity agencies from the Five Eyes alliance issued a joint statement warning that frontier AI models are expected to exceed current industry expectations within months rather than years. The statement said AI would accelerate both defensive and offensive cyber capabilities and that the timeline for such advances is measured in months.

It called for a whole-of-society response, stating that cyber risk is now a core business and leadership responsibility. The warning followed a U.S. decision earlier this month to block foreign nationals from using an Anthropic AI model called Fable.

The same restriction was applied to another Anthropic model, Mythos, which is available only to vetted organizations.

Background on the models Mythos was released earlier this year and is designed to detect vulnerabilities in cyber systems. Fable 5 is described as a more community-friendly version of Mythos. The U.S. government cited national security advice when suspending access for foreign nationals in June.

The Five Eyes statement did not name specific companies or models. Olivia Shen, an expert at the University of Sydney’s United States Studies Centre, said additional advanced models could be developed by other states or companies in the near term. In March the Australian government signed a non-binding memorandum of understanding with Anthropic under its national AI plan, which promotes a light-touch regulatory approach.