The record
Written from the 1 report below. Nothing here is unsourced.
- Anthropic has launched new Compliance API endpoints that give security teams visibility into what Claude Code agents do on developers' local machines.
- Claude Code runs commands, reads files, and connects to third-party tools using the employee's own credentials, and before August 2026 Anthropic's native controls offered limited insight into this activity.
- The new endpoints expose session metadata and transcripts of everything communicated to Anthropic's models.
- However, the report stresses that these logs alone are not enough for governance, since they cannot confirm whether an agent's access is legitimate without endpoint telemetry tying activity to owners, credentials, and permissions.
What to watch next
- How security teams combine the new Compliance API logs with endpoint telemetry for full governance
- Whether Anthropic expands managed-settings policy controls beyond static allow/deny rules
- Adoption of the endpoints by enterprises given 82% of surveyed professionals found unknown AI agents in their environments
Coverage1
1 report
English national1
All filed from India
Named United States · Anthropic · Claude Code · Cloud Security Alliance · Token Security
The 1 report is listed beside the record.
Ask this story
Answers cite the reports above, or say they can't.