{
 "public_id": "AIR-2024-0009",
 "incident_id": "2024-02-microsoft-openai-state-actor-llm-abuse",
 "date_disclosed": "2024-02-14",
 "date_occurred": "2024-02-14",
 "title": "Microsoft and OpenAI disclosed state-sponsored groups from North Korea, Iran, Russia and China using LLMs for target research and spear-phishing",
 "class": "in_the_wild",
 "attack_surface": "other",
 "initial_vector": "policy_violating_use",
 "agent_capability_abused": [
  "content_generation",
  "target_research",
  "code_generation"
 ],
 "target_product": "OpenAI models via API and ChatGPT",
 "target_vendor": "OpenAI / Microsoft",
 "autonomy_level": "assistant",
 "impact_realized": true,
 "impact_type": [
  "ecosystem_abuse",
  "information_disclosure"
 ],
 "blast_radius": "global defence, cybersecurity and cryptocurrency targets; five named state-linked groups including Kimsuky (Emerald Sleet) and Forest Blizzard",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI02"
  ],
  "mitre_atlas": [],
  "labels": [
   "nation_state",
   "threat_intel_disclosure",
   "spear_phishing",
   "account_enforcement"
  ]
 },
 "defense_present": "usage policy enforcement and threat-actor monitoring",
 "defense_bypassed": true,
 "fix_shipped": "accounts terminated; joint Microsoft/OpenAI threat-intel programme established",
 "primary_sources": [
  "https://www.theguardian.com/technology/2024/feb/14/north-korea-iran-ai-hacking-microsoft",
  "https://incidentdatabase.ai/cite/644"
 ],
 "verified_quote": "Microsoft said it detected and disrupted, in collaboration with business partner OpenAI, many threats that used or attempted to exploit AI technology they had developed.",
 "confidence": "high",
 "notes": "The first joint vendor threat-intelligence disclosure of nation-state LLM abuse, and the template every later OpenAI/Anthropic misuse report follows. Use was productivity-augmenting (research, translation, phishing text), not yet autonomous — the baseline that later Cursor- and Claude-assisted intrusion records should be read against. AI Incident Database incident 644.",
 "risk_category": "privacy",
 "risk_secondary": [],
 "risk_subcategory": "Unclassified",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "human_acted_on_output",
 "agency_debatable": false,
 "agency_reason": "state actors used LLM output for research and phishing they sent themselves",
 "vector_family": "misuse",
 "vector_detail": "policy_violating_use",
 "guardrail_kind": [
  "human_supervision",
  "abuse_detection"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [
   644
  ],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI02",
 "agent_framework": "unknown",
 "model_provider": "multiple",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "sensitive_data_exposure",
 "eai_domain": "privacy",
 "eai_secondary": [],
 "eai_reason": "impact rule: sensitive_data_exposure",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI02 (impact=tool_misuse)",
 "submitted_by": "EAI Internal"
}