{
 "public_id": "AIR-2024-0046",
 "incident_id": "2024-11-deepseek-prompt-injection-account-takeover",
 "date_disclosed": "2024-11-29",
 "date_occurred": null,
 "title": "DeepSeek AI: prompt injection to XSS to full account takeover",
 "class": "responsible_disclosure",
 "attack_surface": "consumer_chatbot",
 "initial_vector": "indirect_prompt_injection_document",
 "agent_capability_abused": [
  "document_analysis",
  "html_output_rendering"
 ],
 "target_product": "DeepSeek chat (DeepSeek-R1-Lite)",
 "target_vendor": "DeepSeek",
 "autonomy_level": "assistant",
 "impact_realized": false,
 "impact_type": [
  "account_takeover",
  "session_hijack"
 ],
 "blast_radius": "DeepSeek chat users",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI03"
  ],
  "mitre_atlas": [],
  "labels": [
   "xss",
   "account_takeover",
   "output_rendering",
   "indirect_prompt_injection"
  ]
 },
 "defense_present": "none - unsanitized model output rendered in the app",
 "defense_bypassed": true,
 "fix_shipped": "fixed one day after disclosure via the site's Contact Us feature",
 "primary_sources": [
  "https://embracethered.com/blog/posts/2024/deepseek-ai-prompt-injection-to-xss-and-account-takeover/"
 ],
 "verified_quote": "This post demonstrated how it is possible for a prompt injection to entirely take over a user's account if an application is vulnerable to XSS, which the LLM can exploit.",
 "confidence": "high",
 "notes": "Johann Rehberger. The userToken was read out of localStorage and exfiltrated.",
 "risk_category": "security_fraud",
 "risk_secondary": [],
 "risk_subcategory": "Prompt Injection",
 "risk_sub_secondary": [
  "Unauthorized Access & Account Takeover"
 ],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": true,
 "agency_reason": "the rendered model output executed as XSS and exfiltrated the token",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "indirect_prompt_injection_document",
 "guardrail_kind": [
  "none"
 ],
 "guardrail_outcome": "not_applicable",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "unknown",
 "model_provider": "deepseek",
 "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": [
  "confused_deputy"
 ],
 "eai_reason": "impact rule: privilege_escalation",
 "asi_curated": [
  "ASI01"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI01 | added ASI03 (signal=identity)",
 "submitted_by": "EAI Internal"
}