Anthropic has shipped a new agent identity access model that gives AI agents their own scoped credentials instead of borrowing human employees' logins. The model ties agent permissions to channels rather than people, applies the principle of least privilege, and logs agent actions separately from human actions — solving the security, continuity, and accountability problems that arise when autonomous software impersonates a person.
Key facts
What: Anthropic's new 'agent identity' model lets Claude agents hold their own scoped accounts for tools like GitHub and Slack, tied to channels -- instead of borrowing a human employee's login.
When: 2026-06-24
Primary source: read the source
