{
 "public_id": "AIR-2026-0144",
 "incident_id": "2026-06-uw-agentic-browsers-same-origin",
 "date_disclosed": "2026-06-30",
 "date_occurred": null,
 "title": "Agentic browsers break the same-origin policy: cross-origin data theft demonstrated on ChatGPT Atlas",
 "class": "research_demo",
 "attack_surface": "browser_agent",
 "initial_vector": "indirect_prompt_injection_web_content",
 "agent_capability_abused": [
  "cross_frame_reading",
  "authenticated_browsing",
  "page_summarization"
 ],
 "target_product": "ChatGPT Atlas (Agent Mode); Chrome with Gemini; Claude for Chrome; Perplexity Comet; Brave Leo AI; Microsoft Edge with Copilot; Firefox AI Mode",
 "target_vendor": "OpenAI; Google; Anthropic; Perplexity; Brave; Microsoft; Mozilla",
 "autonomy_level": "semi_autonomous",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "privacy_violation",
  "security_boundary_violation"
 ],
 "blast_radius": "seven agentic browsers tested; four permitted same-origin-policy bypass",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI02"
  ],
  "mitre_atlas": [],
  "labels": [
   "same_origin_policy",
   "agentic_browser",
   "cross_origin_theft"
  ]
 },
 "defense_present": "browser same-origin policy plus each agent's injection defenses",
 "defense_bypassed": true,
 "fix_shipped": "varied; several vendors declined or did not reply",
 "primary_sources": [
  "https://agent-security.cs.washington.edu/agentic_browsers_sop.html",
  "https://seclab.cs.washington.edu/2026/07/01/agentic-browsers-and-the-same-origin-policy/",
  "https://www.washington.edu/news/2026/06/30/some-agentic-ai-browsers-come-with-major-cybersecurity-risks-uw-study-finds/"
 ],
 "verified_quote": "the same-origin policy is reduced to the strength of an agent's defenses against prompt injections in multiple agentic browsers today",
 "confidence": "high",
 "notes": "Franziska Roesner and David Kohlbrenner, UW Security and Privacy Research Lab; project page last updated 15 Apr 2026, UW News release 30 Jun 2026. End-to-end 'successful cross-origin data theft attack' was executed against ChatGPT Atlas in Agent Mode; conditions for similar attacks were found in Chrome with Gemini, Claude for Chrome and Perplexity Comet. Disclosed to all vendors with 60+ days notice: Brave, Google and Microsoft acknowledged; OpenAI and Firefox declined; Anthropic had not replied at publication. NOT the same work as arXiv 2606.14027 'Same-Origin Policy for Agentic Browsers' (Wang, Chen, Li, Song, Gong) - see quarantine.",
 "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 agentic browsers read cross-origin content and exfiltrated it",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "indirect_prompt_injection_web_content",
 "guardrail_kind": [
  "permission_scope"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "unknown",
 "model_provider": "multiple",
 "model_name": "",
 "detection_mechanism": "external_researcher",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "data_exfiltration_channels",
 "eai_domain": "privacy",
 "eai_secondary": [
  "sensitive_data_exposure",
  "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"
}