# AI Agents Run Intrusion Campaigns

Attackers increasingly hand reconnaissance, exploitation and lateral movement to AI agents while humans only pick targets. This raises the number and speed of intrusions against poorly secured cloud and database infrastructure, including the systems that hold crypto keys and run cryptojacking. Agent-assisted breach reports will keep recurring, and defenders will keep being pushed toward automated detection.

- Conviction: 38 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-11-15T00:00:00.000Z
- Last updated: 2026-11-15T14:09:58.038Z
- Canonical: https://polylog.news/crypto/trends/ai-agent-run-intrusions
- Publisher: Polylog

## Recent evidence

- [confirms] Investigators Say AI Agents Helped Attackers Gain Access at 10 of at Least 19 Targeted Organizations (2026-11-15): A November 12 investigators' report counts 25 operations in August and September in which AI agents helped attackers gain access at 10 of at least 19 targeted organizations, with humans picking the targets and able to steer the tooling. Separate researchers detailed agent-run campaigns against Japan and a Monero-mining operation on exposed database servers.
