Japan: AI Agent Policy
Summary: Japan’s Digital Agency has opened council-level work (22 October 2026) on how existing law treats AI agents that express intent, represent users and conclude contracts, tied to permissions, authentication and audit records.
Sources: 2026-10-25-ai-agent-identity-news.md
Last updated: 2026-10-25
2026-10-22: AI and Digital Reform Promotion Council, first meeting
- A draft issues paper proposes clarifying how existing law applies to agents expressing intent, representing users and concluding contracts — including contracts exceeding the user’s intention. It links the legal analysis to permissions, authentication and logs, and proposes system-level safeguards: agent harnesses, identity authentication, guardrails, authority management, evaluation and audit records (source: 2026-10-25-ai-agent-identity-news.md, citing Material 4, pp. 4–6 and council page; primary).
- A member submission (Emiko Maruyama) examines attribution of agent actions, allocation of unexpected losses and the records needed to identify responsibility (source: 2026-10-25-ai-agent-identity-news.md, citing Material 9; primary).
- Status: proposals for deliberation — not enacted agent-personhood rules or binding identity requirements.
Japan was previously named only in passing as a comparator in the OECD working-group article (see estonia-ai-agent-id); this is the first Japanese primary source in the wiki.