Symantec honeypot captured an autonomous LLMjacking agent allowed 100,000 unattended steps
What happened
Symantec captured the instruction set of an autonomous offensive agent after it contacted a honeypot posing as an exposed AI service. The agent hijacks exposed AI models to scan, exploit and mine, and was set to run up to 100,000 actions without human approval. Symantec reports no confirmed victims.
The operators sent their working prompt to the decoy, expecting a live model. Layered instructions told the model to override its safety controls, disguised credential theft as legitimate development work and allowed it to run up to 100,000 malicious actions without human approval. The tool's modules cover scanning, exploitation and mining, and an obfuscated script installed Monero-mining software, hid it and erased logs. The honeypot did not execute the attacker's tools. Symantec captured the instruction set in summer 2026 and recorded more than 127,000 external requests from tens of thousands of addresses.
How it happened
AI-run attack: Autonomous intrusion. The attacker's own AI agent carried out the break-in, working through the steps with little human involvement.
Sources do not say whose AI model was involved. The product type was AI agent.
Impact
Data was exposed. Data involved: credentials. There is no evidence that anyone actually accessed the data.
Severity is rated Low, calculated from how far the incident reached and how sensitive the data was.
Was the AI at fault?
Unclear. Sources do not establish whether the model's own behaviour or the surrounding systems were to blame.
Status and timeline
Confirmed by the affected organization, the AI vendor, a regulator or a named security research firm. This was a research finding: no real victims are known.
It became public on 21 Oct 2026.
Symantec published its findings on 21 October. The sources describe no response from AI providers.