AI Agent Fleets Use Public Maps

Researchers observed a large 'fleet' of independent AI agents running on cloud infrastructure systematically querying a public mapping API for directions to specific entrances of parks, hospitals and other sites. They were detected through urlquery logs (a domain‑scanning proxy that agents use when they cannot fetch sites directly), underscoring how easily persistent agent activity can be audited — and how easily it can be used for large‑scale physical‑world reconnaissance. — Automated agent fleets probing public map services create new privacy, security, and attribution problems — they can enable scalable reconnaissance, inform physical targeting or surveillance, and complicate regulation of cross‑cloud agent activity.