CVE-2025-54132
📋 TL;DR
This vulnerability in Cursor code editor versions below 1.3 allows attackers to exfiltrate sensitive information via Mermaid diagram image rendering. Attackers can embed images that fetch data to external servers after successful prompt injection. Users of Cursor versions below 1.3 who process untrusted data are affected.
💻 Affected Systems
- Cursor code editor
📦 What is this software?
Cursor by Anysphere
⚠️ Risk & Real-World Impact
Worst Case
Complete exfiltration of sensitive code, credentials, or proprietary information from the user's environment to attacker-controlled servers.
Likely Case
Limited data leakage from chat sessions or processed files containing sensitive information.
If Mitigated
No data exfiltration occurs due to prompt injection prevention and network controls.
🎯 Exploit Status
Requires successful prompt injection first, then Mermaid image rendering triggers the exfiltration.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: 1.3 and above
Vendor Advisory: https://github.com/cursor/cursor/security/advisories/GHSA-43wj-mwcc-x93p
Restart Required: No
Instructions:
1. Open Cursor editor 2. Go to Settings > About 3. Check for updates 4. Install version 1.3 or higher 5. Restart Cursor if prompted
🔧 Temporary Workarounds
Disable Mermaid diagram rendering
allPrevent Mermaid diagrams from rendering in chat to block the exfiltration vector
Network egress filtering
allBlock outbound connections to unknown external domains from Cursor process
🧯 If You Can't Patch
- Avoid processing untrusted data or code in Cursor chat
- Use network monitoring to detect unusual outbound connections from Cursor
🔍 How to Verify
Check if Vulnerable:
Check Cursor version in Settings > About. If version is below 1.3, you are vulnerable.Check Version:
In Cursor: Open Settings > About to view versionVerify Fix Applied:
Confirm version is 1.3 or higher in Settings > About and test that Mermaid diagrams with external images no longer trigger network requests.📡 Detection & Monitoring
Log Indicators:
- Unusual outbound HTTP requests from Cursor process
- Mermaid diagram rendering errors
Network Indicators:
- Outbound connections to unknown domains from Cursor process on ports 80/443
- Image fetch requests to external servers
SIEM Query:
process_name:"Cursor" AND (destination_port:80 OR destination_port:443) AND NOT destination_domain IN [allowed_domains]