Symantec reports capturing an offensive agent's operating instructions through an AI honeypot in summer 2026. The request described using exposed models to drive attacks and mining, with up to 100,000 unattended steps; the honeypot executed none of its tools and the report does not demonstrate a successful intrusion by this agent.
Why it matters
It documents attacker interest in stealing inference access, while separating intended autonomy from observed execution.
Key facts
As stated in the sources, with where to find them.
- The request included about 74 offensive tool definitions. Its 100,000-step allowance is an operator instruction, not an observed run length.Tooling and tradecraft; Autonomous operation and hidden mining
- Three earlier tool results supplied inside the request found no flaw. The honeypot returned a generic fallback and the session ended; it did not validate that supplied history independently.Tool calls recorded in the request
- The more than 127,000 external requests refer to overall honeypot telemetry, not successful compromises or actions by this single agent.What the attackers were looking for
Findings that cite this record
No tracked finding cites this record yet.
Key questions this bears on
- How are attackers using AI agents in real operations?To run parts of intrusions at scale: vendors report a state-aligned actor and criminal groups using open-source agent frameworks and commercial models.
Sources
Related records
Oct 29, 2026
Oct 29, 2026
Oct 29, 2026
Oct 21, 2026
Oct 21, 2026
Oct 21, 2026