{
 "public_id": "AIR-2026-0266",
 "incident_id": "2026-09-mcp-python-sdk-oauth-credential-theft",
 "date_disclosed": "2026-09-29",
 "date_occurred": null,
 "title": "Official MCP Python SDK: rogue MCP server can obtain OAuth client credentials through missing issuer validation",
 "class": "responsible_disclosure",
 "attack_surface": "mcp_server",
 "initial_vector": "malicious_mcp_server_oauth_discovery",
 "agent_capability_abused": [
  "oauth_discovery",
  "mcp_client_authentication",
  "token_endpoint_selection"
 ],
 "target_product": "MCP Python SDK",
 "target_vendor": "Model Context Protocol (Anthropic-stewarded open source)",
 "autonomy_level": "assistant",
 "impact_realized": false,
 "impact_type": [
  "credential_theft",
  "unauthorized_access"
 ],
 "blast_radius": "applications using the SDK as an MCP client over HTTP with OAuth: versions 1.9.1 to 1.29.1 and 2.0.0 to 2.1.1",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI04",
   "ASI03"
  ],
  "mitre_atlas": [],
  "labels": [
   "oauth",
   "missing_issuer_validation",
   "rogue_mcp_server"
  ]
 },
 "defense_present": null,
 "defense_bypassed": true,
 "fix_shipped": "MCP Python SDK 1.30.0 and 2.2.0 (issuer checks listed in release notes as behavior changes on 2026-09-07)",
 "primary_sources": [
  "https://thehackernews.com/2026/09/official-mcp-python-sdk-flaw-can-let.html",
  "https://github.com/modelcontextprotocol/python-sdk/security/advisories/GHSA-qx49-fqc8-xw99"
 ],
 "verified_quote": "trick an application built on the official MCP Python SDK into handing over the OAuth credentials it uses to log in to a real service",
 "confidence": "high",
 "notes": "Advisory GHSA-qx49-fqc8-xw99, published 2026-09-28 alongside the Cycode write-up; no CVE seen. The quote is from the news coverage of the advisory, which is the source carrying it; the advisory is listed as the authoritative record. Cycode demonstrated the full credential exchange in a test. Neither the advisory nor Cycode reports attacks using the flaw. The fix first shipped quietly in the 1.30.0 and 2.2.0 release notes, ahead of the advisory.",
 "risk_category": "security_fraud",
 "risk_secondary": [],
 "risk_subcategory": "Credential & Secret Theft",
 "risk_sub_secondary": [
  "Unauthorized Access & Account Takeover"
 ],
 "ai_entity": "generative",
 "agency": "ai_as_target",
 "agency_debatable": false,
 "agency_reason": "flaw in the SDK's OAuth handling lets a rogue MCP server capture credentials; no model action in the chain",
 "vector_family": "malicious_component",
 "vector_detail": "malicious_mcp_server_oauth_discovery",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI04",
 "agent_framework": "mcp_server",
 "model_provider": "anthropic",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "credential_theft",
 "eai_domain": "access_control_and_permissions",
 "eai_secondary": [
  "privilege_escalation"
 ],
 "eai_reason": "subcategory: Credential & Secret Theft",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI04 (vector_family=malicious_component); ASI03 (eai_category=credential_theft)",
 "submitted_by": "EAI Internal"
}