{
 "public_id": "AIR-2026-0131",
 "incident_id": "2026-06-mozilla-tabstack-injection",
 "date_disclosed": "2026-06-08",
 "date_occurred": null,
 "title": "Brave: hidden white-on-white instructions hijack Mozilla Tabstack's cloud browser agent to exfiltrate conversation history",
 "class": "responsible_disclosure",
 "attack_surface": "browser_agent",
 "initial_vector": "indirect_prompt_injection_web_page",
 "agent_capability_abused": [
  "page_summarization",
  "navigation",
  "form_submission",
  "conversation_history_access"
 ],
 "target_product": "Mozilla Tabstack (cloud-hosted browser agent); Cotypist (on-device macOS autocomplete)",
 "target_vendor": "Mozilla; Cotypist",
 "autonomy_level": "semi_autonomous",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "credential_exposure"
 ],
 "blast_radius": "Tabstack users summarizing attacker-controlled pages; Cotypist users opening attacker-supplied local documents",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI02"
  ],
  "mitre_atlas": [],
  "labels": [
   "hidden_text",
   "white_on_white",
   "conversation_history_theft",
   "on_device_agent",
   "local_document_injection"
  ]
 },
 "defense_present": "none effective",
 "defense_bypassed": true,
 "fix_shipped": "unknown",
 "primary_sources": [
  "https://brave.com/blog/indirect-prompt-injection/"
 ],
 "verified_quote": "The agent never summarized the page. Instead, it executed all three injected steps in sequence.",
 "confidence": "high",
 "notes": "Two 2026 cases in one Brave writeup, showing the class is not confined to Comet/Atlas. (1) Mozilla Tabstack: asked to summarize a page, the cloud agent instead followed white-on-white instructions, navigated to an attacker domain, pulled its own conversation history and POSTed it via a form. (2) Cotypist: a fully on-device macOS autocomplete model was steered by instructions inside local documents into suggesting false content and potentially surfacing credentials inline - Brave notes small local models are at higher risk of failing to separate instructions from data.",
 "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": "the cloud browser agent navigated and submitted the conversation history",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "indirect_prompt_injection_web_page",
 "guardrail_kind": [
  "none"
 ],
 "guardrail_outcome": "not_applicable",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "unknown",
 "model_provider": "unknown",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "data_exfiltration_channels",
 "eai_domain": "privacy",
 "eai_secondary": [
  "credential_theft",
  "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 ASI02 (impact=tool_misuse)",
 "submitted_by": "EAI Internal"
}