{
 "public_id": "AIR-2026-0047",
 "incident_id": "2026-02-trailofbits-comet-audit",
 "date_disclosed": "2026-02-20",
 "date_occurred": null,
 "title": "Threat modeling and prompt injection audit of Perplexity Comet extracts private Gmail data",
 "class": "responsible_disclosure",
 "attack_surface": "browser_agent",
 "initial_vector": "indirect_prompt_injection_web_page",
 "agent_capability_abused": [
  "page_summarization",
  "authenticated_session_reuse",
  "gmail_access"
 ],
 "target_product": "Perplexity Comet",
 "target_vendor": "Perplexity",
 "autonomy_level": "semi_autonomous",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration"
 ],
 "blast_radius": "Comet users summarizing attacker-controlled pages while signed into Gmail",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI03"
  ],
  "mitre_atlas": [],
  "labels": [
   "browser_agent",
   "prompt_injection",
   "paid_audit"
  ]
 },
 "defense_present": null,
 "defense_bypassed": true,
 "fix_shipped": "Perplexity shipped updates before launch",
 "primary_sources": [
  "https://blog.trailofbits.com/2026/02/20/using-threat-modeling-and-prompt-injection-to-audit-comet/"
 ],
 "verified_quote": "The vulnerabilities we found reflect how AI agents behave when external content isn't treated as untrusted input.",
 "confidence": "high",
 "notes": "Four techniques: summarization instructions, fake security mechanisms, fake system instructions, fake user request. Perplexity security lead Kyle Polley is quoted in the post. This was a commissioned audit, not an unsolicited disclosure.",
 "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": "Comet reused the Gmail session and extracted private mail",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "indirect_prompt_injection_web_page",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "unknown",
 "model_provider": "unknown",
 "model_name": "",
 "detection_mechanism": "external_researcher",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "data_exfiltration_channels",
 "eai_domain": "privacy",
 "eai_secondary": [
  "confused_deputy"
 ],
 "eai_reason": "subcategory: Data Exfiltration via Context",
 "asi_curated": [
  "ASI01"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI01 | added ASI03 (signal=identity)",
 "submitted_by": "EAI Internal"
}