{
 "id": "AIR-2026-0357",
 "slug": "zombieagent-indirect-prompt-injection-on-chatgpt",
 "title": "ZombieAgent Indirect Prompt Injection on ChatGPT",
 "summary": "Researchers demonstrated an indirect prompt injection attack abusing ChatGPT connectors and Deep Research capabilities. Concealed instructions within external content were executed during normal user tasks, highlighting connector risks.",
 "type": "attack_on_ai",
 "categories": [
  "prompt_injection",
  "agent_hijack"
 ],
 "severity": 1,
 "targets": [
  "OpenAI"
 ],
 "products": [
  "ChatGPT",
  "ChatGPT Deep Research"
 ],
 "incident_date": "2025-09-25",
 "incident_date_precision": "day",
 "first_reported": "2026-07-31",
 "sources": [
  {
   "url": "https://atlas.mitre.org/studies/AML.CS0066",
   "publisher": "MITRE ATLAS",
   "title": "ZombieAgent: Data Exfiltration Attack on ChatGPT",
   "published_at": "2026-07-31"
  },
  {
   "url": "https://radware.com/blog/threat-intelligence/zombieagent",
   "publisher": "radware.com",
   "title": "ZombieAgent: New ChatGPT Vulnerabilities Let Data Theft Continue (and Spread)",
   "published_at": "2026-07-31"
  }
 ],
 "source_count": 2,
 "confidence": 0.95,
 "status": "published",
 "origin": "seed_atlas",
 "tags": [
  "zombieagent",
  "prompt-injection",
  "chatgpt",
  "indirect-prompt-injection"
 ],
 "created_at": "2026-10-18T06:55:08.239Z",
 "updated_at": "2026-10-18T06:55:08.239Z"
}