Disclosed vulnerability
Cross Plugin Request Forgery: ChatGPT as a confused deputy between plugins
AIR-2023-0011 AIR record · 2023-05-openai-cross-plugin-request-forgery
From the supporting source
Random webpages and data will hijack your AI, steal your stuff and spend your money.embracethered.com
The record
Disclosed
2023-05-28
Vendor
OpenAI
Product
ChatGPT Plugins (including Zapier)
Risk domain
Agent behaviour › Unsafe actuation
Also touches
Data exfiltration channels, Sensitive data exposure, Confused deputy
System type
Generative model
Who acted
Agent acted — ChatGPT invoked a second plugin as a confused deputy
Attack surface
Skill / plugin
Autonomy level
Assistant
Initial vector
Indirect prompt injection web content
Impact type
Data exfiltration, Pii disclosure, Unauthorized action
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI01 – Agent Goal Hijack
- MITRE ATLAS
- AML.T0048 – External Harms
- NIST AI RMF
- MEASURE 2.6; MANAGE 1.3
- EU AI Act
- Annex III §1(c)
- ISO/IEC
- TR 24028; 24029-1; 23894
- AIUC-1
- C. Safety
- Scope
- agent misuse
- Mitigation
- Require confirmation for destructive actions; Implement dry-run modes; Use layered safety controls.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
ChatGPT plugin users
Capability abused
Defense & remediation
Recorded defense
none initially; Zapier later added authenticated confirmation for sensitive operations
Guardrail kinds
None
Guardrail outcome
Not applicable
Fix shipped
partially mitigated by Zapier; no generic OpenAI-side fix at publication
Taxonomy & references
Curator notes
Johann Rehberger. Coins Cross Plugin Request Forgery (CPRF), the direct ancestor of today's cross-MCP confused-deputy attacks.
Supporting sources
Cite as
Enkrypt AI. "Cross Plugin Request Forgery: ChatGPT as a confused deputy between plugins". AIR, AIR-2023-0011. https://agentincident.vercel.app/i/AIR-2023-0011/ (accessed 2026-10-19).