{
 "id": "AIR-2025-0129",
 "slug": "openai-chatgpt-models-jailbroken-to-provide-hazardous-weapon",
 "title": "OpenAI ChatGPT Models Jailbroken to Provide Hazardous Weapons Instructions",
 "summary": "Investigation revealed that multiple OpenAI models could be jailbroken to bypass safety guardrails and generate hazardous weapons instructions. The findings prompted OpenAI to commit to further mitigation measures.",
 "type": "attack_on_ai",
 "categories": [
  "jailbreak"
 ],
 "severity": 1,
 "targets": [
  "OpenAI"
 ],
 "products": [
  "ChatGPT",
  "o4-mini",
  "GPT-5-mini"
 ],
 "incident_date": "2025-10-10",
 "incident_date_precision": "day",
 "first_reported": "2025-10-10",
 "sources": [
  {
   "url": "https://incidentdatabase.ai/cite/1238",
   "publisher": "AI Incident Database",
   "title": "OpenAI ChatGPT Models Reportedly Jailbroken to Provide Chemical, Biological, and Nuclear Weapons Instructions",
   "published_at": "2025-10-10"
  }
 ],
 "source_count": 1,
 "confidence": 0.95,
 "status": "published",
 "origin": "seed_aiid",
 "tags": [
  "jailbreak",
  "safety-bypass",
  "llm"
 ],
 "created_at": "2026-10-18T06:33:31.967Z",
 "updated_at": "2026-10-18T06:59:35.797Z"
}