Japan's AI Basic Plan revision names Claude Mythos as the advanced-threat benchmark
Japan's government revised its AI Basic Plan just six months after its December 2025 formulation, explicitly naming Anthropic's Claude Mythos as the reference model for advanced AI cybersecurity risk — the first time Japan has anchored national policy to a specific frontier model. The revision commits to continuously updating AI law as capabilities advance and adds international cooperation with foreign governments and AI developers as a new pillar, shifting Japan from static guidelines to a framework that tracks the frontier.
Source: nippon.com ↗
The government reviewed the basic plan, which serves as a guideline for its AI policy, just about six months after its formulation last December, in light of rapid advancements in AI technology.
Japanese government
Why this matters
- → Japan anchors national AI policy to specific frontier model for first time.
- → Six-month revision cycle signals shift from static rules to capability-tracking framework.
- → International cooperation pillar addresses cross-border AI risks and developer accountability.
Benchmarking the frontier