The West may have a head start in using AI for cyber defense, but it may be a short one. Intelligence officials from the Five Eyes alliance say the advantage enjoyed by the US, UK, Canada, New Zealand, and Australia is measured in months, not years, as rivals race to weaponize the same technology for attacks that are faster, more adaptive, and harder to pin down.
That warning lands at an awkward moment for the industry. Anthropic recently had to restrict access to its advanced models, Mythos 5 and Fable 5, after US authorities barred the company from giving them to foreign governments. Instead of building a neat geopolitical firewall, the company ended up cutting off most users, including many in the US, because selective access proved too difficult to enforce.
Five Eyes says the AI cyber defense gap is already narrow
The broader message from the intelligence community is blunt: frontier models are outperforming industry expectations on both offense and defense. That means governments and companies are no longer dealing with the old, mostly manual world of phishing and malware kits; they are facing systems that can help attackers pick targets, adjust tactics, and spread effort across many targets at once.
For now, the alliance says Western countries still lead in folding AI into commercial and military security workflows. But if that lead is truly only a few months long, then the real race is not about who gets the best model first. It is about who can wire it into corporate defenses quickly enough to matter.
Google has already flagged AI-assisted attacks
That concern is not coming out of nowhere. In May, Google said it suspected geopolitical rivals of using AI in attacks on information infrastructure, with algorithms that adapt automatically to changing conditions. The detail that should make security teams nervous is not just automation, but flexibility: an attack that learns as it goes is much harder to blunt with static defenses.
- Five Eyes includes the US, UK, Canada, New Zealand, and Australia.
- The alliance says the Western lead in AI security is only a matter of months.
- Anthropic’s Mythos 5 and Fable 5 were restricted after US government pressure.
The next fight is inside company networks
There is a second, quieter shift here: the best defense may no longer come from buying one magical platform, but from companies using their own AI systems against incoming attacks. Western intelligence officials are explicitly urging businesses to lean harder on in-house resources, which is a polite way of saying the old perimeter model is looking tired.
That will reward firms that can move fast and punish the rest. If AI is already making attacks more distributed and more difficult to trace, then the companies that wait for a perfect security playbook may end up writing incident reports instead.