AI-orchestrated intrusion incidents
An intrusion in which an AI agent carried out several attack steps, such as reconnaissance, exploitation and lateral movement, with people directing it.
- Incidents on this page
- 4
- Confirmed by a primary document (E1)
- 4
- Verified losses
- None yetNo record here has a verified loss figure
- Latest disclosure
- 2026
About these records
These records cover intrusions where attackers handed whole stages of the work to an AI agent rather than asking a chatbot for help with one task. The agent scans, tests flaws, writes code, moves between systems and reports back, while human operators choose targets and approve key steps. Most of what is known comes from AI vendors who found their own products being misused, so each record states who made the assessment and how much of the work the vendor says the AI did.
The registry holds 4 records of this type. All of them are confirmed in a primary document (evidence level E1). 3 are rated severe or critical on the five-point scale. The latest was disclosed on 10 September 2026.
Records
Newest by date of occurrence.
| Record | Occurred | Incident | Severity |
|---|---|---|---|
| AHL-2026-0002 | Dec 2025 to Aug 2026 | GTG-50014: suspected ShinyHunters affiliates used Claude | 5 |
| AHL-2026-0001 | Dec 2025 to Aug 2026 | GTG-20006: Russian espionage automated with Claude | 4 |
| AHL-2025-0001 | Sep 2025 | GTG-1002: Claude Code used to run a state espionage campaign | 4 |
| AHL-2025-0002 | 2025 | GTG-2002: Claude Code used for data theft and extortion | 3 |
By month
By month of occurrence. Records dated only by year are not shown.
Show as table
| Month | Incidents |
|---|---|
| Nov 2026 | 0 |
| Oct 2026 | 0 |
| Sep 2026 | 0 |
| Aug 2026 | 0 |
| Jul 2026 | 0 |
| Jun 2026 | 0 |
| May 2026 | 0 |
| Apr 2026 | 0 |
| Mar 2026 | 0 |
| Feb 2026 | 0 |
| Jan 2026 | 0 |
| Dec 2025 | 2 |
| Nov 2025 | 0 |
| Oct 2025 | 0 |
| Sep 2025 | 1 |