Anthropic gives Claude AI agents their own credentials instead of borrowing human logins

Anthropic has introduced a new agent identity access model, announced on June 24, 2026, that assigns Claude AI agents their own scoped accounts for tools like GitHub and Slack. Previously, AI agents operated by borrowing a human employee's login credentials, creating security, continuity, and accountability problems. The new model ties agent permissions to specific channels rather than individuals, applying the principle of least privilege so agents access only what their task requires. Agent actions are now logged separately from human actions, making audits clearer and access revocation straightforward. This infrastructure change is seen as a critical prerequisite for companies to deploy AI agents at scale in a secure and auditable way.