{
 "public_id": "AIR-2025-0021",
 "incident_id": "2024-09-akirabot-llm-spam-at-scale",
 "date_disclosed": "2025-04-09",
 "date_occurred": "2024-09-01",
 "title": "AkiraBot used OpenAI's API plus CAPTCHA bypass to autonomously spam 80,000+ website contact forms and chat widgets",
 "class": "in_the_wild",
 "attack_surface": "autonomous_ops",
 "initial_vector": "misuse_of_llm_api_by_operator",
 "agent_capability_abused": [
  "content_generation",
  "form_submission",
  "captcha_bypass",
  "proxy_rotation"
 ],
 "target_product": "OpenAI chat API (abused) via the AkiraBot framework",
 "target_vendor": "OpenAI (abused)",
 "autonomy_level": "fully_autonomous",
 "impact_realized": true,
 "impact_type": [
  "spam_at_scale",
  "abuse"
 ],
 "blast_radius": "more than 400,000 domains targeted; at least 80,000 successfully spammed",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI10"
  ],
  "mitre_atlas": [],
  "labels": [
   "llm_generated_spam",
   "captcha_bypass",
   "personalised_at_scale"
  ]
 },
 "defense_present": "site CAPTCHAs and anti-spam filters",
 "defense_bypassed": true,
 "fix_shipped": "OpenAI disabled the abused API key after disclosure",
 "primary_sources": [
  "https://www.sentinelone.com/labs/akirabot-ai-powered-bot-bypasses-captchas-spams-websites-at-scale/",
  "https://incidentdatabase.ai/cite/1365"
 ],
 "verified_quote": "SentinelLABS assesses that AkiraBot has targeted more than 400,000 websites and successfully spammed at least 80,000 websites since September 2024.",
 "confidence": "high",
 "notes": "AI Incident Database incident 1365; SentinelLABS research. Included because the bot autonomously browsed, solved CAPTCHAs and submitted forms - not merely generated text.",
 "risk_category": "product_misuse",
 "risk_secondary": [],
 "risk_subcategory": "Capability Abuse",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "the bot browsed, solved CAPTCHAs and submitted 80,000 forms itself",
 "vector_family": "misuse",
 "vector_detail": "misuse_of_llm_api_by_operator",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [
   1365
  ],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI10",
 "agent_framework": "unknown",
 "model_provider": "openai",
 "model_name": "",
 "detection_mechanism": "threat_intel",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "policy_drift",
 "eai_domain": "governance",
 "eai_secondary": [],
 "eai_reason": "subcategory: Capability Abuse",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI10 (adversary-operated agent (vector_family=misuse))",
 "submitted_by": "EAI Internal"
}