{
 "public_id": "AIR-2025-0188",
 "incident_id": "2025-12-unit42-mcp-sampling-attacks",
 "date_disclosed": "2025-12-05",
 "date_occurred": null,
 "title": "Prompt injection attack vectors through MCP sampling: conversation hijacking and covert tool invocation",
 "class": "research_demo",
 "attack_surface": "mcp_server",
 "initial_vector": "malicious_mcp_server",
 "agent_capability_abused": [
  "mcp_sampling",
  "tool_invocation",
  "file_system_access"
 ],
 "target_product": "MCP clients supporting sampling (an unnamed coding copilot and an unnamed code editor)",
 "target_vendor": null,
 "autonomy_level": "copilot_with_approval",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "resource_theft",
  "persistence"
 ],
 "blast_radius": "Any MCP client implementing the sampling feature",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI04",
   "ASI03"
  ],
  "mitre_atlas": [],
  "labels": [
   "mcp",
   "sampling",
   "conversation_hijacking",
   "covert_tool_invocation"
  ]
 },
 "defense_present": null,
 "defense_bypassed": true,
 "fix_shipped": null,
 "primary_sources": [
  "https://unit42.paloaltonetworks.com/model-context-protocol-attack-vectors/"
 ],
 "verified_quote": "MCP sampling relies on an implicit trust model and lacks robust, built-in security controls.",
 "confidence": "high",
 "notes": "Authors Huang, Rao, Li, Ji, Hu. Three vectors: resource theft, conversation hijacking, covert tool invocation. Unit 42 deliberately withheld vendor names.",
 "risk_category": "security_fraud",
 "risk_secondary": [
  "privacy"
 ],
 "risk_subcategory": "Prompt Injection",
 "risk_sub_secondary": [
  "Persistence & Covert C2"
 ],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "MCP sampling drove covert tool invocation and exfiltration",
 "vector_family": "malicious_component",
 "vector_detail": "malicious_mcp_server",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "mcp_server",
 "model_provider": "microsoft",
 "model_name": "",
 "detection_mechanism": "external_researcher",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "privilege_escalation",
 "eai_domain": "access_control_and_permissions",
 "eai_secondary": [
  "data_exfiltration_channels"
 ],
 "eai_reason": "impact rule: privilege_escalation",
 "asi_curated": [
  "ASI01"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI01 | added ASI04 (vector_family=malicious_component), ASI03 (eai_category=privilege_escalation, no code execution)",
 "submitted_by": "EAI Internal"
}