Disclosed vulnerability
Codex for Everything: email-triage injection exfiltrates legal, financial and security correspondence to an attacker's Google Form
AIR-2026-0119 AIR record · 2026-05-openai-codex-for-everything-email-exfil
From the supporting source
This vulnerability was responsibly disclosed on Apr 21, 2026, and the vulnerability has been remediated by OpenAI.promptarmor.com
The record
Disclosed
2026-05-21
Vendor
OpenAI
Product
Codex for Everything
Risk domain
Privacy › Data exfiltration channels
Also touches
Confused deputy
System type
Generative model
Who acted
Agent acted — Codex read the mail and submitted it to the attacker's Google Form
Attack surface
Enterprise assistant
Autonomy level
Semi-autonomous
Initial vector
Indirect prompt injection email
Impact type
Data exfiltration
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI02 – Tool Misuse and Exploitation
- MITRE ATLAS
- AML.T0024 – Exfiltration via AI Inference API
- NIST AI RMF
- MAP 4.2
- EU AI Act
- Article 10
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- A. Data & Privacy
- Scope
- tool misuse
- Mitigation
- Monitor network traffic for anomalous patterns; Restrict egress destinations.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
Codex for Everything users with an email plugin connected
Capability abused
Defense & remediation
Recorded defense
unspecified
Guardrail kinds
Unknown
Guardrail outcome
Silent
Fix shipped
remediated by OpenAI before public disclosure
Taxonomy & references
Curator notes
PromptArmor. Reported 2026-04-21, public 2026-05-21. Covers OpenAI's extension of Codex beyond coding to general day-to-day tasks with connectors. 'No user interaction is required beyond submission of the initial email triage query.'
Supporting sources
Cite as
Enkrypt AI. "Codex for Everything: email-triage injection exfiltrates legal, financial and security correspondence to an attacker's Google Form". AIR, AIR-2026-0119. https://agentincident.vercel.app/i/AIR-2026-0119/ (accessed 2026-10-19).