{
 "public_id": "AIR-2026-0044",
 "incident_id": "2026-02-checkpoint-ai-in-the-middle-c2",
 "date_disclosed": "2026-02-17",
 "date_occurred": null,
 "title": "AI in the Middle: using Grok and Microsoft Copilot web-fetch features as covert C2 proxies",
 "class": "research_demo",
 "attack_surface": "enterprise_assistant",
 "initial_vector": "abuse_of_unauthenticated_web_fetch",
 "agent_capability_abused": [
  "url_fetch",
  "browsing",
  "summarisation"
 ],
 "target_product": "Grok and Microsoft Copilot web interfaces (abused)",
 "target_vendor": "xAI, Microsoft",
 "autonomy_level": "assistant",
 "impact_realized": false,
 "impact_type": [
  "covert_c2",
  "data_exfiltration",
  "detection_evasion"
 ],
 "blast_radius": "enterprises that allow AI service domains as trusted egress",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI03"
  ],
  "mitre_atlas": [
   "AML.CS0061"
  ],
  "labels": [
   "ai_service_as_c2",
   "trusted_domain_egress",
   "no_credential_to_revoke"
  ]
 },
 "defense_present": "egress filtering on suspicious infrastructure",
 "defense_bypassed": true,
 "fix_shipped": null,
 "primary_sources": [
  "https://research.checkpoint.com/2026/ai-in-the-middle-turning-web-based-ai-services-into-c2-proxies-the-future-of-ai-driven-attacks/",
  "https://raw.githubusercontent.com/mitre-atlas/atlas-data/main/dist/v6/ATLAS-2026.07.yaml"
 ],
 "verified_quote": "This technique was demonstrated against platforms such as Grok and Microsoft Copilot, leveraging anonymous web access combined with browsing and summarization prompts",
 "confidence": "high",
 "notes": "Check Point Research. Because the abused web-fetch flows need no authentication, defenders cannot kill the channel by rotating keys or suspending accounts.",
 "risk_category": "security_fraud",
 "risk_secondary": [
  "privacy"
 ],
 "risk_subcategory": "Persistence & Covert C2",
 "risk_sub_secondary": [],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "Grok and Copilot fetched the attacker's URLs, acting as the C2 relay",
 "vector_family": "conventional_vulnerability",
 "vector_detail": "abuse_of_unauthenticated_web_fetch",
 "guardrail_kind": [
  "permission_scope"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [
   "AML.CS0061"
  ],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI03",
 "agent_framework": "unknown",
 "model_provider": "multiple",
 "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",
  "policy_drift"
 ],
 "eai_reason": "subcategory: Persistence & Covert C2",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI03 (eai_category=privilege_escalation, no code execution)",
 "submitted_by": "EAI Internal"
}