Capture Notes

This paper argues that communication graph metadata in autonomous agent interoperability is an integrity risk, not only a privacy risk.

AI security relevance:

• Shows that who contacts whom, when, and how often can reveal workflow class and enable adversarial timing or preemption.
• Relevant to A2A/MCP multi-agent deployments where content encryption may not hide capability-labeled workflow structure.
• Provides a new gap for graph-level privacy and workflow integrity controls.

Suggested ingest focus:

• Link with MAS non-human identity, workflow provenance, and multi-agent protocol security topics.