Five Eyes Warn Frontier AI Hacking Capabilities Are 'Months Away' from Public Availability

The Five Eyes intelligence alliance—comprising the United States, Canada, the United Kingdom, Australia, and New Zealand—has issued a stark joint statement warning that advanced frontier AI models with sophisticated hacking capabilities are 'months away' from being publicly available. The statement, signed by senior officials including the NSA's Director of the Cybersecurity Directorate David Imbordino and acting CISA Director Nick Andersen, asserts that models like Anthropic's Fable 5 and OpenAI's Daybreak will soon be accessible to the public despite efforts by AI companies to restrict access. 'The timeline is not years, it is months,' the agencies wrote, emphasizing that the rapid pace of AI development will fundamentally transform both offensive and defensive cyber operations.

The warning highlights that AI models capable of exploiting cybersecurity weaknesses are already available through multiple channels, including older commercial models, open-source versions, and foreign or black-market sources. The agencies point to legacy systems, slow patching cycles, unnecessary internet connectivity, weak identity and access controls, and a lack of pre-incident planning as key vulnerabilities that AI will excel at exploiting. The statement notes that the capabilities described in a recent Amazon threat intelligence report—which convinced the Trump administration to impose export controls on Fable 5—can already be achieved with older models like Claude Opus and Claude Sonnet, as well as open-source Chinese models.

Anthropic has shut down access to its Fable 5 and Mythos 5 models following the export controls, though the company maintains that the White House decision was based on a 'misunderstanding.' The dispute remains unresolved. Meanwhile, open-source models now lag behind frontier AI by only 6-8 months, meaning that restricted capabilities quickly become widely available. The agencies stress that 'the rapid pace of frontier AI development means cyber risk assumptions can become outdated in months, not years,' urging organizations to act proactively rather than reactively.

Programs like Anthropic's Project Glasswing and OpenAI's Trusted Access for Cyber Program aim to give defenders a head start by providing AI systems for cyber defense, helping organizations find and fix vulnerabilities before malicious actors can exploit them. However, the recommended guidance from the Five Eyes remains largely unchanged from decades of cybersecurity best practices: governments and businesses must stop treating digital security as an afterthought or compliance issue. 'Success will come from getting the basics right, acting quickly, and integrating cyber security into core business strategy,' the agencies wrote. 'Those that do not will face growing operational and strategic disadvantage.'

The statement underscores that traditional risk assumptions are becoming obsolete at an alarming rate. As AI models grow more powerful, the window for patching vulnerabilities and securing systems shrinks. The agencies call for a fundamental shift in mindset, urging organizations to prepare for a landscape where AI-driven attacks are routine and automated. The warning aligns with growing concerns from cybersecurity experts about the dual-use nature of AI, where the same models that can defend networks can also be weaponized for offensive operations.

This joint advisory from the Five Eyes is one of the most explicit government warnings to date about the near-term threat posed by advanced AI. It signals that intelligence agencies expect the democratization of hacking capabilities to accelerate, potentially outpacing traditional defense mechanisms. The message is clear: organizations must prioritize foundational cybersecurity practices now, or face severe consequences as AI-powered threats become ubiquitous.