{
 "license": "CC-BY-4.0",
 "attribution": "Fide AI, Agentic Cyber Explorer",
 "url": "https://agentic-cyber-explorer.pages.dev/events/symantec-llmjacking-agent-honeypot-capture-2026/",
 "asOf": "2026-10-30",
 "id": "symantec-llmjacking-agent-honeypot-capture-2026",
 "date": "2026-10-21",
 "datePrecision": "day",
 "title": "Symantec captures an offensive agent request at a simulated exposed AI endpoint",
 "lane": "attack",
 "kind": "misuse-report",
 "summary": "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.",
 "whyItMatters": "It documents attacker interest in stealing inference access, while separating intended autonomy from observed execution.",
 "actors": [
  "broadcom"
 ],
 "topics": [
  "ai-enabled-intrusion",
  "threat-intelligence",
  "access-controls"
 ],
 "atlas": [
  "access-gate",
  "tools"
 ],
 "artifacts": [],
 "sources": [
  {
   "url": "https://www.security.com/threat-intelligence/ai-llmjacking-skynet",
   "publisher": "Symantec / Carbon Black Threat Hunter Team",
   "title": "Autonomous LLMjacking: ‘Skynet’ Agent Hijacks Exposed AI Models to Run Attacks",
   "date": "2026-10-21",
   "type": "primary",
   "accessed": "2026-10-24"
  }
 ],
 "keyFacts": [
  {
   "fact": "The request included about 74 offensive tool definitions. Its 100,000-step allowance is an operator instruction, not an observed run length.",
   "locator": "Tooling and tradecraft; Autonomous operation and hidden mining"
  },
  {
   "fact": "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.",
   "locator": "Tool calls recorded in the request"
  },
  {
   "fact": "The more than 127,000 external requests refer to overall honeypot telemetry, not successful compromises or actions by this single agent.",
   "locator": "What the attackers were looking for"
  }
 ],
 "significance": 3,
 "fideQuestions": [],
 "methods": [
  "agent-orchestrated-intrusion"
 ],
 "review": "assistant-drafted",
 "addedOn": "2026-10-24",
 "revisions": []
}