{
 "public_id": "AIR-2026-0128",
 "incident_id": "2026-06-anthropic-mitre-attack-ai-threats",
 "date_disclosed": "2026-06-03",
 "date_occurred": null,
 "title": "Anthropic study of 832 banned threat-actor accounts: AI use concentrated in malware development and lateral movement",
 "class": "in_the_wild",
 "attack_surface": "coding_agent",
 "initial_vector": "aggregate_analysis_of_banned_accounts",
 "agent_capability_abused": [
  "code_generation",
  "lateral_movement_assistance",
  "account_discovery"
 ],
 "target_product": "Claude",
 "target_vendor": "Anthropic",
 "autonomy_level": "assistant",
 "impact_realized": true,
 "impact_type": [
  "capability_uplift"
 ],
 "blast_radius": "832 banned accounts studied",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI10"
  ],
  "mitre_atlas": [],
  "labels": [
   "mitre_atlas_referenced",
   "measurement",
   "threat_landscape",
   "taxonomy_gap"
  ]
 },
 "defense_present": null,
 "defense_bypassed": null,
 "fix_shipped": null,
 "primary_sources": [
  "https://www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack"
 ],
 "verified_quote": "writing malware (560 of the 832 accounts we studied, or 67.3%, used AI for this purpose)",
 "confidence": "high",
 "notes": "Useful as a base-rate record rather than a single incident: it is the largest published denominator on how threat actors actually use a frontier agent. Anthropic argues MITRE ATT&CK under-counts AI-enabled operations - the Nov 2025 GTG-1002 case mapped to only 30 techniques across 13 tactics, comparable to medium-risk actors.",
 "risk_category": "security_fraud",
 "risk_secondary": [],
 "risk_subcategory": "Over-permissioned execution & lateral movement",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "human_acted_on_output",
 "agency_debatable": true,
 "agency_reason": "an aggregate study of humans using AI for uplift in their own operations",
 "vector_family": "other",
 "vector_detail": "aggregate_analysis_of_banned_accounts",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI10",
 "agent_framework": "unknown",
 "model_provider": "anthropic",
 "model_name": "",
 "detection_mechanism": "vendor_disclosure",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "privilege_escalation",
 "eai_domain": "access_control_and_permissions",
 "eai_secondary": [],
 "eai_reason": "subcategory: Over-permissioned execution & lateral movement",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "override",
 "asi_reason": "override: An umbrella threat-intelligence record about 832 banned accounts using Claude for malware and intrusion work. The identity rule fired on the legacy privilege_escalation category, but nothing here turns on delegation or credential inheritance: it is adversary-operated agent use, the same reading applied to the vector_family=misuse records.",
 "submitted_by": "EAI Internal"
}