{
 "public_id": "AIR-2026-0019",
 "incident_id": "2026-01-pillar-operation-bizarre-bazaar",
 "date_disclosed": "2026-01-28",
 "date_occurred": "2025-12",
 "title": "Operation Bizarre Bazaar: first attributed LLMjacking campaign reselling stolen access to exposed LLM and MCP endpoints through a commercial marketplace",
 "class": "in_the_wild",
 "attack_surface": "mcp_server",
 "initial_vector": "internet_scanning_of_exposed_endpoints",
 "agent_capability_abused": [
  "unauthenticated_inference_api",
  "mcp_server_tool_access",
  "lateral_movement_via_mcp"
 ],
 "target_product": "exposed Ollama (port 11434), unauthenticated vLLM / OpenAI-compatible APIs (port 8000), open MCP servers",
 "target_vendor": "victim organizations (self-hosted)",
 "autonomy_level": "semi_autonomous",
 "impact_realized": true,
 "impact_type": [
  "resource_theft",
  "data_exfiltration",
  "lateral_movement"
 ],
 "blast_radius": "35,000 attack sessions captured Dec 2025 - Jan 2026, averaging 972 attacks/day",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI06",
   "ASI04",
   "ASI03"
  ],
  "mitre_atlas": [],
  "labels": [
   "llmjacking",
   "criminal_marketplace",
   "shodan_censys_scanning",
   "mcp_as_pivot",
   "attributed_actor"
  ]
 },
 "defense_present": "none - endpoints exposed without authentication",
 "defense_bypassed": true,
 "fix_shipped": "n/a",
 "primary_sources": [
  "https://www.pillar.security/blog/operation-bizarre-bazaar-first-attributed-llmjacking-campaign-with-commercial-marketplace-monetization"
 ],
 "verified_quote": "Attackers systematically target Ollama instances running on port 11434 without authentication, OpenAI-compatible APIs on port 8000 exposed to the internet",
 "confidence": "high",
 "notes": "Pillar Security. Actor alias 'Hecker' (aka Sakuya, LiveGamer101) running silver.inc, 'The Unified LLM API Gateway', on bulletproof hosting with Discord/Telegram resale. Three-role structure: scanner bots, validators, marketplace operator; exposed endpoints validated and enumerated within 2-8 hours. Agentic relevance: compromised MCP servers bridge to internal databases and cloud APIs, so this is a lateral-movement vector, not just compute theft.",
 "risk_category": "security_fraud",
 "risk_secondary": [
  "privacy"
 ],
 "risk_subcategory": "Supply Chain Compromise",
 "risk_sub_secondary": [
  "Over-permissioned execution & lateral movement"
 ],
 "ai_entity": "generative",
 "agency": "ai_as_target",
 "agency_debatable": true,
 "agency_reason": "exposed LLM and MCP endpoints scanned and resold",
 "vector_family": "conventional_vulnerability",
 "vector_detail": "internet_scanning_of_exposed_endpoints",
 "guardrail_kind": [
  "none"
 ],
 "guardrail_outcome": "not_applicable",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI06",
 "agent_framework": "mcp_server",
 "model_provider": "multiple",
 "model_name": "",
 "detection_mechanism": "external_researcher",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "supply_chain_vulnerabilities",
 "eai_domain": "tool_misuse",
 "eai_secondary": [
  "privilege_escalation",
  "data_exfiltration_channels"
 ],
 "eai_reason": "subcategory: Supply Chain Compromise",
 "asi_curated": [
  "ASI06"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI06 | added ASI04 (eai_category=supply_chain_vulnerabilities), ASI03 (signal=identity)",
 "submitted_by": "EAI Internal"
}