Japan is exploring the use of cutting-edge artificial intelligence in its preemptive cyber defence strategy to counter advanced AI-driven threats, including models like Anthropic's Claude Mythos.

National Cyber Director Yoichi Iida warns that the window between vulnerability discovery and exploitation could shrink dramatically, even to fractions of previous times, underscoring the need for faster, automated yet human-guided defence.

Iida emphasizes advancing automation in cyber defence while preserving human involvement in final decision-making, though concrete steps have not been laid out.

In March, Prime Minister Sanae Takaichi’s Cabinet approved enabling active cyber defence from October 1 to preempt serious attacks on critical infrastructure and livelihoods.

Anthropic's Claude Mythos is regarded as highly adept at finding security vulnerabilities, with access initially limited to select IT firms and financial institutions due to misuse concerns.

Iida stresses not relying on a single AI model and cites the U.S. export controls on Mythos-class models in June as a cautionary example for diversifying AI tools.

If adversaries exploit cutting-edge AI, nation-sponsored cyberattacks could become far more sophisticated and intense.

Iida asserts that Japan will inevitably deploy advanced AI technology to bolster cyber defenses.