Ledger / Jul 2026
APT passes an AI agent off as an authorized pentest against a Southeast Asian organization
Strong
Trend states without hedging that "In one case an AI agent ran its own reconnaissance and lateral movement", in a summary report that shows no evidence. Trend is a security firm, not the model provider, and a firm's one-line statement with no evidence shown is Strong.
- Model
- Not namedAI agents (model not named)
- Agent harness
- Nonechat or tool use only
- Attributed country
- Not attributedthe source names no country
- Sectors hit
- enterprise broadAn organisation in Southeast Asia
An actor, possibly Earth Lamia, convinced an autonomous AI agent the operation was an authorized penetration test; the agent then ran scanning, exploitation, credential harvesting and password spraying against a Southeast Asian organization.
“In one case tied to a possible Earth Lamia operation, they jailbroke an autonomous AI agent that then conducted its own network scanning, exploitation, credential harvesting, and password spraying against a target in Southeast Asia.”Trend Micro
Sources
- Date
- Jul 2026
- Actor
- Unattributed APT (possibly Earth Lamia, per Trend Micro) (unknown), as attributed by Trend Micro
- Nexus
- China (possible, per Trend Micro)
- Targets
- An organisation in Southeast Asia
- What the AI did
- Agentic intrusion, Recon, Exploit dev
- Model as named
- AI agents (model not named)
- Model basis
- A researcher names it
- Autonomy
- Semi-autonomous
- Outcome
- Scanning, exploitation, credential harvesting and spraying against the organization.