{
 "public_id": "AIR-2025-0175",
 "incident_id": "2025-11-microsoft-copilot-actions-windows-xpia",
 "date_disclosed": "2025-11-13",
 "date_occurred": null,
 "title": "Microsoft ships Copilot Actions on Windows 11 off-by-default, documenting that cross-prompt injection can cause data exfiltration or malware installation",
 "class": "responsible_disclosure",
 "attack_surface": "computer_use_agent",
 "initial_vector": "cross_prompt_injection_from_documents_and_ui_elements",
 "agent_capability_abused": [
  "local_file_access",
  "gui_control_click_type_scroll",
  "app_automation"
 ],
 "target_product": "Copilot Actions on Windows 11 / agent workspace",
 "target_vendor": "Microsoft",
 "autonomy_level": "semi_autonomous",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "malware_installation"
 ],
 "blast_radius": "Windows 11 Insider machines where an administrator enables Experimental agentic features",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI06",
   "ASI05"
  ],
  "mitre_atlas": [],
  "labels": [
   "vendor_acknowledged_risk",
   "xpia",
   "off_by_default",
   "agent_account_isolation",
   "os_level_agent"
  ]
 },
 "defense_present": "disabled by default; separate standard 'agent account'; contained agent workspace as a defended security boundary; ACL-limited access to Documents/Downloads/Desktop/Pictures; user approval for sensitive actions; signed-agent requirement",
 "defense_bypassed": false,
 "fix_shipped": "n/a - shipped as an experimental preview with the risk documented",
 "primary_sources": [
  "https://learn.microsoft.com/en-us/windows/security/book/operating-system-agentic-security",
  "https://support.microsoft.com/en-US/Windows/Ai/Ai-Features/experimental-agentic-features"
 ],
 "verified_quote": "agentic AI applications introduce novel security risks, such as cross-prompt injection (XPIA), where malicious content embedded in UI elements or documents can override agent instructions, leading to unintended actions like data exfiltration or malware installation.",
 "confidence": "high",
 "notes": "Vendor threat disclosure rather than an observed exploit: retained as the OS-level computer-use-agent baseline because Microsoft documents the concrete threat model for its shipped preview. No public exploit of Copilot Actions on Windows was found during research for this record; that remains an open lead. Microsoft additionally tells users to 'only enable this feature if you understand the security implications'. Doc dated 2025-11-13, support article updated 2025-12-05.",
 "risk_category": "security_fraud",
 "risk_secondary": [
  "privacy"
 ],
 "risk_subcategory": "Prompt Injection",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": true,
 "agency_reason": "the documented failure is Copilot Actions clicking and typing under injection",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "cross_prompt_injection_from_documents_and_ui_elements",
 "guardrail_kind": [
  "human_approval",
  "permission_scope",
  "sandbox_isolation"
 ],
 "guardrail_outcome": "not_enabled",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "windows_copilot_actions",
 "model_provider": "microsoft",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "data_exfiltration_channels",
 "eai_domain": "privacy",
 "eai_secondary": [],
 "eai_reason": "impact rule: data_exfiltration_channels",
 "asi_curated": [
  "ASI01",
  "ASI06"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI01,ASI06 | added ASI05 (impact=code_exec)",
 "submitted_by": "EAI Internal"
}