Disclosed vulnerability
Cursor IDE: arbitrary data exfiltration via Mermaid diagram image rendering (CVE-2025-54132)
AIR-2025-0077 AIR record · 2025-08-cursor-mermaid-exfiltration
From the supporting source
An attacker can use Mermaid to exfiltrate sensitive information to a third-party attacker controlled server when performing a prompt injection.embracethered.com
The record
Disclosed
2025-08-04
Vendor
Anysphere / Cursor
Product
Cursor IDE
Risk domain
Privacy › Data exfiltration channels
Also touches
Credential theft, Confused deputy
System type
Generative model
Who acted
Agent acted — Cursor rendered the Mermaid diagram, fetching the attacker URL
Attack surface
Coding agent
Autonomy level
Semi-autonomous
Initial vector
Indirect prompt injection source code
Impact type
Data exfiltration, Credential theft
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
Cursor users before v1.3
Capability abused
Defense & remediation
Recorded defense
Cursor sandbox and image domain restrictions, bypassed via Mermaid's external image loading
Guardrail kinds
Sandbox isolation
Guardrail outcome
Silent
Fix shipped
resolved 2025-07-07, shipped in Cursor 1.3 on 2025-07-29
Taxonomy & references
Curator notes
Month of AI Bugs 2025. | Merged from duplicate id `2025-08-cursor-mermaid-exfil`. Also reported as: 'Cursor IDE: arbitrary data exfiltration via Mermaid diagram rendering (CVE-2025-54132)'.
Supporting sources
Cite as
Enkrypt AI. "Cursor IDE: arbitrary data exfiltration via Mermaid diagram image rendering (CVE-2025-54132)". AIR, AIR-2025-0077. https://agentincident.vercel.app/i/AIR-2025-0077/ (accessed 2026-10-19).