{
 "public_id": "AIR-2026-0207",
 "incident_id": "2026-08-zenity-claude-sandbox-dns-shell",
 "date_disclosed": "2026-08-27",
 "date_occurred": null,
 "title": "DNS egress in Claude's sandbox: data exfiltration and a bidirectional DNS shell, regressed after the first fix",
 "class": "responsible_disclosure",
 "attack_surface": "computer_use_agent",
 "initial_vector": "sandbox_egress_channel",
 "agent_capability_abused": [
  "code_execution_sandbox",
  "dns_resolution"
 ],
 "target_product": "Claude sandbox environment",
 "target_vendor": "Anthropic",
 "autonomy_level": "semi_autonomous",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "command_and_control"
 ],
 "blast_radius": "Claude sandboxed code-execution environments behind the egress proxy",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI05"
  ],
  "mitre_atlas": [],
  "labels": [
   "dns_tunneling",
   "egress_filter_bypass",
   "sandbox_escape_channel",
   "regression"
  ]
 },
 "defense_present": "Egress proxy blocking outbound HTTP/HTTPS",
 "defense_bypassed": true,
 "fix_shipped": "Initial DNS exfiltration fixed December 2025; regressed in a later deployment and re-reported",
 "primary_sources": [
  "https://labs.zenity.io/post/it-s-always-dns-in-claude-s-sandbox-from-data-exfiltration-to-a-bidirectional-dns-shell"
 ],
 "verified_quote": "DNS resolution happens before those filters can intervene.",
 "confidence": "high",
 "notes": "Author Dmitry Lozovoy. Anthropic treated the bidirectional shell as a duplicate of the underlying DNS channel issue and paid one bounty for both. Notable as a documented security regression in an agent sandbox.",
 "risk_category": "security_fraud",
 "risk_secondary": [
  "privacy"
 ],
 "risk_subcategory": "Remote Code Execution & Sandbox Escape",
 "risk_sub_secondary": [
  "Persistence & Covert C2",
  "Over-permissioned execution & lateral movement"
 ],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "the agent's sandbox resolved DNS to exfiltrate and hold a shell",
 "vector_family": "other",
 "vector_detail": "sandbox_egress_channel",
 "guardrail_kind": [
  "permission_scope"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI05",
 "agent_framework": "unknown",
 "model_provider": "anthropic",
 "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": "subcategory: Remote Code Execution & Sandbox Escape",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI05 (vector=code_exec)",
 "submitted_by": "EAI Internal"
}