{
 "public_id": "AIR-2026-0132",
 "incident_id": "2026-06-microsoft-searchleak",
 "date_disclosed": "2026-06-09",
 "date_occurred": null,
 "title": "SearchLeak: one-click exfiltration of mail, MFA codes and files from Microsoft 365 Copilot Enterprise Search",
 "class": "responsible_disclosure",
 "attack_surface": "enterprise_assistant",
 "initial_vector": "url_parameter_prompt_injection",
 "agent_capability_abused": [
  "enterprise_search",
  "mailbox_retrieval",
  "html_rendering",
  "image_fetch"
 ],
 "target_product": "Microsoft 365 Copilot (Enterprise Search)",
 "target_vendor": "Microsoft",
 "autonomy_level": "assistant",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "mfa_bypass"
 ],
 "blast_radius": "Microsoft 365 Copilot Enterprise tenants; victim only had to click a genuine microsoft.com link",
 "cve": [
  "CVE-2026-42824"
 ],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI02"
  ],
  "mitre_atlas": [],
  "labels": [
   "one_click",
   "parameter_to_prompt",
   "race_condition",
   "csp_bypass",
   "ssrf",
   "trusted_domain_abuse"
  ]
 },
 "defense_present": "Copilot output sanitizer and Content Security Policy",
 "defense_bypassed": true,
 "fix_shipped": "Microsoft remediated server-side under CVE-2026-42824 (max severity critical); no customer action required",
 "primary_sources": [
  "https://www.varonis.com/blog/searchleak"
 ],
 "verified_quote": "CSP Bypass via Bing SSRF: Bing's image-search endpoint, allowlisted in the Content Security Policy, performs a server-side fetch to an attacker-controlled URL.",
 "confidence": "high",
 "notes": "Varonis Threat Labs. Three chained flaws: (1) Parameter-to-Prompt injection where the URL q parameter in Copilot Enterprise Search is passed to Copilot as an executable prompt, (2) an HTML rendering race condition where an <img> tag fires before the output sanitizer, (3) a CSP bypass via Bing SSRF. Proof of concept only; no observed exploitation. CVSS reporting is inconsistent across sources (Microsoft 6.5, NVD 7.5, Tenable 7.8) despite the 'critical' label; treat severity claims with care. date_disclosed reflects the June 2026 publication window; the exact day was not verified.",
 "risk_category": "privacy",
 "risk_secondary": [],
 "risk_subcategory": "Data Exfiltration via Context",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "Copilot searched the mailbox and fetched the exfiltration image",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "url_parameter_prompt_injection",
 "guardrail_kind": [
  "output_filter"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [
   "CVE-2026-42824"
  ],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "m365_copilot",
 "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": "subcategory: Data Exfiltration via Context",
 "asi_curated": [
  "ASI01"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI01 | added ASI02 (impact=tool_misuse)",
 "submitted_by": "EAI Internal"
}