{
 "public_id": "AIR-2025-0009",
 "incident_id": "2025-02-openai-peer-review-surveillance",
 "date_disclosed": "2025-02-21",
 "date_occurred": null,
 "title": "'Peer Review': China-origin accounts used ChatGPT to build and pitch a social media surveillance tool",
 "class": "in_the_wild",
 "attack_surface": "consumer_chatbot",
 "initial_vector": "direct_misuse_for_surveillance_tooling",
 "agent_capability_abused": [
  "code_debugging",
  "document_analysis",
  "sales_material_generation"
 ],
 "target_product": "ChatGPT (tool itself ran on non-OpenAI models)",
 "target_vendor": "OpenAI",
 "autonomy_level": "assistant",
 "impact_realized": true,
 "impact_type": [
  "surveillance_tooling",
  "human_rights_harm"
 ],
 "blast_radius": "Tool claimed to feed real-time reports on Western protests, including Uyghur rights protests, to Chinese security services",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI09"
  ],
  "mitre_atlas": [],
  "labels": [
   "surveillance",
   "authoritarian_misuse",
   "multi_model"
  ]
 },
 "defense_present": "Usage policies prohibiting communications surveillance",
 "defense_bypassed": true,
 "fix_shipped": "Accounts banned (Feb 2025)",
 "primary_sources": [
  "https://cdn.openai.com/threat-intelligence-reports/disrupting-malicious-uses-of-our-models-february-2025-update.pdf"
 ],
 "verified_quote": "They used our models to generate detailed descriptions, consistent with sales pitches, of a social media listening tool that they claimed to have used to feed real-time reports about protests in the West to the Chinese security services.",
 "confidence": "high",
 "notes": "impact_realized is set true on the basis of the operators' own claim of deployment, which OpenAI could not independently verify - a reasonable reader could set it false. Notable multi-model pattern: OpenAI models were used to build and sell a tool that ran on a different (Llama-derived) model, so no single provider sees the whole operation. OpenAI's Oct 2025 report describes continued, similar PRC-linked activity.",
 "risk_category": "bias_fairness",
 "risk_secondary": [
  "product_misuse"
 ],
 "risk_subcategory": "Wrongful Automated Decision & Rights Harm",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "human_acted_on_output",
 "agency_debatable": false,
 "agency_reason": "operators used ChatGPT to build and pitch the surveillance tool themselves",
 "vector_family": "misuse",
 "vector_detail": "direct_misuse_for_surveillance_tooling",
 "guardrail_kind": [
  "abuse_detection"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI09",
 "agent_framework": "unknown",
 "model_provider": "openai",
 "model_name": "",
 "detection_mechanism": "vendor_disclosure",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "human_manipulation",
 "eai_domain": "agent_behaviour",
 "eai_secondary": [],
 "eai_reason": "subcategory: Wrongful Automated Decision & Rights Harm",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI09 (eai_category=human_manipulation)",
 "submitted_by": "EAI Internal"
}