{
 "public_id": "AIR-2025-0012",
 "incident_id": "2025-02-truffle-security-live-api-keys-in-common-crawl",
 "date_disclosed": "2025-02-27",
 "date_occurred": "2024-12",
 "title": "Truffle Security found 11,908 live credentials inside a Common Crawl archive used as LLM training data",
 "class": "research_demo",
 "attack_surface": "other",
 "initial_vector": "training_data_contamination",
 "agent_capability_abused": [
  "training_data_ingestion",
  "code_suggestion"
 ],
 "target_product": "Common Crawl (December 2024 archive) and models trained on it",
 "target_vendor": "Common Crawl; downstream model developers",
 "autonomy_level": "assistant",
 "impact_realized": true,
 "impact_type": [
  "information_disclosure",
  "credential_exposure"
 ],
 "blast_radius": "11,908 credentials that authenticated successfully in a 400 TB crawl; affected vendors collectively rotated or revoked several thousand keys",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI02"
  ],
  "mitre_atlas": [],
  "labels": [
   "training_data_contamination",
   "credential_exposure",
   "insecure_pattern_reinforcement",
   "corpus_hygiene"
  ]
 },
 "defense_present": "none; the crawl had no secret scanning",
 "defense_bypassed": false,
 "fix_shipped": "Truffle Security contacted affected vendors and helped rotate or revoke several thousand keys",
 "primary_sources": [
  "https://trufflesecurity.com/blog/research-finds-12-000-live-api-keys-and-passwords-in-deepseek-s-training-data",
  "https://thehackernews.com/2025/02/12000-api-keys-and-passwords-found-in.html",
  "https://incidentdatabase.ai/cite/956/"
 ],
 "verified_quote": "11,908 Live Secrets were detected using TruffleHog in 400TB of web data. (Note: A ‘live’ secret authenticates successfully.)",
 "confidence": "high",
 "notes": "Class is research_demo because this was a researcher-led scan of a public corpus, not disclosure of a shipped-product vulnerability. impact_realized remains true because the credentials authenticated and vendors rotated or revoked several thousand keys. The LLM relevance is a retained ai_as_target boundary: Common Crawl is used as training data, but the study does not establish that a model emitted any secret. AI Incident Database incident 956.",
 "risk_category": "privacy",
 "risk_secondary": [],
 "risk_subcategory": "Membership Inference",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "ai_as_target",
 "agency_debatable": true,
 "agency_reason": "live credentials sat in a public training corpus; no AI action in the harm chain",
 "vector_family": "other",
 "vector_detail": "training_data_contamination",
 "guardrail_kind": [
  "none"
 ],
 "guardrail_outcome": "not_applicable",
 "xrefs": {
  "aiid": [
   956
  ],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI02",
 "agent_framework": "unknown",
 "model_provider": "unknown",
 "model_name": "",
 "detection_mechanism": "external_researcher",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "sensitive_data_exposure",
 "eai_domain": "privacy",
 "eai_secondary": [
  "credential_theft"
 ],
 "eai_reason": "subcategory: Membership Inference",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI02 (impact=tool_misuse)",
 "submitted_by": "EAI Internal"
}