{
 "public_id": "AIR-2024-0034",
 "incident_id": "2024-08-microsoft-copilot-bernklau-defamation",
 "date_disclosed": "2024-08-16",
 "date_occurred": "2024-08",
 "title": "Microsoft Copilot told users that court reporter Martin Bernklau was a child abuser and fraudster, conflating him with defendants he had covered, and published his address",
 "class": "safety_failure",
 "attack_surface": "consumer_chatbot",
 "initial_vector": "none_autonomous_action",
 "agent_capability_abused": [
  "retrieval",
  "web_grounding",
  "entity_resolution"
 ],
 "target_product": "Microsoft Copilot",
 "target_vendor": "Microsoft",
 "autonomy_level": "assistant",
 "impact_realized": true,
 "impact_type": [
  "reputational_harm",
  "privacy_violation"
 ],
 "blast_radius": "one named journalist; the failure mode reaches every reporter, lawyer and judge whose name co-occurs with the crimes they document",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI03"
  ],
  "mitre_atlas": [],
  "labels": [
   "no_adversary",
   "entity_conflation",
   "retrieval_grounding_failure",
   "defamation",
   "remediation_failed"
  ]
 },
 "defense_present": "manual suppression of the specific entries",
 "defense_bypassed": true,
 "fix_shipped": "partial; Microsoft's removals were temporary and the output reappeared",
 "primary_sources": [
  "https://www.theregister.com/2024/08/26/microsoft_bing_copilot_ai_halluciation/",
  "https://incidentdatabase.ai/cite/770"
 ],
 "verified_quote": "Copilot falsely claimed Bernklau had been charged with and convicted of child abuse and exploiting dependents.",
 "confidence": "high",
 "notes": "Included despite being text output because it turns on retrieval and grounding in a deployed assistant, and because the remediation failure is the point: Microsoft suppressed the specific claims and the model regenerated them. There is no delete for an inference. Bernklau's contact details were emitted alongside the accusations. AI Incident Database incident 770.",
 "risk_category": "privacy",
 "risk_secondary": [
  "legal_ip",
  "safety_harm"
 ],
 "risk_subcategory": "Unclassified",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": true,
 "agency_reason": "the assistant published defamatory statements and a home address to real users",
 "vector_family": "none_autonomous",
 "vector_detail": "none_autonomous_action",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [
   770
  ],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI03",
 "agent_framework": "unknown",
 "model_provider": "microsoft",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "sensitive_data_exposure",
 "eai_domain": "privacy",
 "eai_secondary": [
  "human_manipulation"
 ],
 "eai_reason": "impact rule: sensitive_data_exposure",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI03 (eai fallback: sensitive_data_exposure)",
 "submitted_by": "EAI Internal"
}