{
 "public_id": "AIR-2026-0251",
 "incident_id": "2026-09-us-military-chatbot-false-chinese-ship-cargo-report",
 "date_disclosed": "2026-09-18",
 "date_occurred": null,
 "title": "A US special-operations analyst's AI chatbot fused open-source and signals intelligence into a false report that a Chinese ship carried nuclear-weapons components; interception aircraft were airborne before the report was rechecked and the operation called off",
 "class": "safety_failure",
 "attack_surface": "enterprise_assistant",
 "initial_vector": "model_hallucination",
 "agent_capability_abused": [
  "retrieval",
  "multi_source_fusion",
  "report_generation"
 ],
 "target_product": "Unidentified AI chatbot used by US Special Operations Command Pacific analysts (commercial or government product not established)",
 "target_vendor": "Unknown",
 "autonomy_level": "assistant",
 "impact_realized": false,
 "impact_type": [
  "false_intelligence",
  "near_miss_military_escalation"
 ],
 "blast_radius": "one Chinese vessel and its crew targeted for interception; US military assets mobilised; operation aborted before contact",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI09"
  ],
  "mitre_atlas": [],
  "labels": [
   "no_adversary",
   "human_acted_on_output",
   "retrieval_grounded_hallucination",
   "automation_bias",
   "military_targeting",
   "near_miss",
   "caught_before_harm",
   "classified_data_fusion"
  ]
 },
 "defense_present": "human review of the intelligence report before the operation",
 "defense_bypassed": false,
 "fix_shipped": "none disclosed; sources say there is no single standard for verifying AI-generated intelligence across the military",
 "primary_sources": [
  "https://edition.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-ship",
  "https://incidentdatabase.ai/cite/1701"
 ],
 "verified_quote": "The bot fused together open-source intelligence with secret signals intelligence in government holdings and reached its fateful conclusion about the material the ship was carrying.",
 "confidence": "medium",
 "notes": "Scope test applied: admitted as a human-acted-on-output event because the consequential output depended on agent-specific retrieval - CNN reports the bot fused OSINT with signals intelligence 'in government holdings' (retrieval over classified stores), and the analyst then used AI again to package the result into a standard intelligence report that was disseminated. A consequential action followed (interception planned, armed personnel preparing to board, military aircraft airborne) and was called off only when officials rechecked the report. impact_realized false: the operation was aborted before any contact, so no party suffered a realized consequence; a source's 'almost started a war' is recorded as a near miss. CNN exclusive (18 Sep 2026) based on four anonymous sources; the chatbot is not identified and it is 'not clear whether the chatbot was a commercially available one or a US government product'; exact date undisclosed ('this spring', during the war with Iran); SOCPAC and the Pentagon did not respond. Single anonymous-sourced press account, hence medium. AI Incident Database incident 1701.",
 "risk_category": "reliability",
 "risk_secondary": [
  "safety_harm"
 ],
 "risk_subcategory": "Hallucination & Factual Inaccuracy",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "human_acted_on_output",
 "agency_debatable": false,
 "agency_reason": "analysts acted on the chatbot's fused report; the system took no action itself",
 "vector_family": "none_autonomous",
 "vector_detail": "model_hallucination",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [
   1701
  ],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI09",
 "agent_framework": "unknown",
 "model_provider": "unknown",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "hallucination",
 "eai_domain": "agent_output_quality",
 "eai_secondary": [],
 "eai_reason": "subcategory: Hallucination & Factual Inaccuracy",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI09 (eai_category=hallucination + reliance signal)",
 "submitted_by": "EAI Internal"
}