Security Analysis

Is Cursor Safe?

An objective security analysis of Cursor AI code editor. What you need to know about its security features, vulnerabilities, and how to use it safely.

Get Starter Scan

Check if code generated by Cursor has security issues.

The Short Answer

Cursor is reasonably safe for most development work when used with precautions. It has SOC 2 certification and Privacy Mode for sensitive work. However, it has had security vulnerabilities, sends code to the cloud, and disables some VS Code security features. Enable Privacy Mode and follow the safety guidelines below for the best protection.

Security Features Assessment

Privacy Mode
Optional mode that prevents your code from being stored or used for training
Must be explicitly enabled in settings
SOC 2 Type II Certification
Enterprise-grade security compliance certification
Demonstrates commitment to security practices
Code Storage
Code is sent to AI for processing but not stored long-term with Privacy Mode
Without Privacy Mode, code may be retained
Local Processing Option
No option to run AI models locally
All AI processing requires cloud connectivity
Workspace Trust
Workspace Trust is disabled by default
Unlike VS Code, doesn't warn about untrusted folders

Known Vulnerabilities

CVE-2025-54135critical
CurXecute - Slack-to-RCE via MCP
Attackers could embed malicious instructions in Slack messages that, when processed by Cursor, execute arbitrary code.
Patched
January 2025
CVE-2025-54136critical
MCPoison - Persistent Backdoor
MCP servers could be compromised to maintain persistent access across team members sharing the same configuration.
Patched
January 2025
CVE-2025-59944high
Case-Sensitivity File Access Bypass
File protection using .cursorignore could be bypassed using different casing (e.g., .ENV vs .env).
Patched
January 2025
Workspace Trustmedium
Disabled Workspace Trust
Cursor disables VS Code's Workspace Trust feature, allowing automatic code execution in untrusted projects.
By Design
Ongoing

Keep Cursor updated to ensure you have all security patches.

How to Use Cursor Safely

Enable Privacy Mode
Go to Settings > Privacy and enable Privacy Mode to prevent code storage
Prevents code from being stored or used for training
Keep Cursor Updated
Enable auto-updates or regularly check for new versions
Ensures you have patches for known vulnerabilities
Use .cursorignore
Add sensitive files like .env, credentials, and private keys to .cursorignore
Prevents AI from accessing sensitive files
Review MCP Servers Carefully
Only install MCP servers from trusted sources, review their code first
Prevents supply chain attacks via malicious MCP servers
Don't Open Untrusted Projects
Be cautious opening projects from unknown sources—Cursor may execute code automatically
Prevents malicious code execution
Scan Generated Code
Review and scan AI-generated code for security issues before deployment
Catches vulnerabilities AI introduces

Code Generated by Cursor

The Bigger Risk: Your Generated Code

While Cursor itself is reasonably secure, the code it generates often has security vulnerabilities. AI-generated code commonly includes:

  • Missing database access controls (RLS not configured)
  • Hardcoded API keys and secrets
  • Client-side only authentication
  • Missing input validation

Scan your Cursor-generated code before deployment to catch these issues.

Get Starter Scanlication

VAS scans applications built with Cursor for security vulnerabilities. Find the issues AI introduced before attackers do.

Get Starter Scan

Frequently Asked Questions

Is Cursor safe to use?

Cursor is generally safe for most use cases when used with precautions. It has SOC 2 Type II certification and offers Privacy Mode. However, it has had security vulnerabilities (now patched), sends code to the cloud for AI processing, and disables Workspace Trust. For sensitive codebases, enable Privacy Mode and follow security best practices.

Does Cursor store my code?

By default, Cursor may retain code for processing and potentially training. With Privacy Mode enabled, Cursor states that code is not stored or used for training. For maximum privacy, enable Privacy Mode in settings before working on sensitive projects.

Is Cursor safe for enterprise use?

Cursor has SOC 2 Type II certification, which is a positive indicator for enterprise use. However, enterprises should evaluate: code processing in the cloud, the need for Privacy Mode, MCP server risks, and the disabled Workspace Trust. Many enterprises use Cursor with appropriate policies in place.

What security vulnerabilities has Cursor had?

Cursor has had several CVEs including: CVE-2025-54135 and CVE-2025-54136 (MCP-related remote code execution), CVE-2025-59944 (case-sensitivity bypass for file protection). All have been patched. Additionally, Workspace Trust is disabled by design, which is a security concern for opening untrusted projects.

Is Cursor safer than GitHub Copilot?

Both tools have similar privacy considerations—code is sent to the cloud for AI processing. Cursor offers Privacy Mode; Copilot has similar enterprise controls. Cursor has had more publicized CVEs recently, but this may reflect more security research rather than inherently worse security. Both require careful configuration for sensitive work.

How do I make Cursor more secure?

1) Enable Privacy Mode in settings, 2) Keep Cursor updated, 3) Use .cursorignore for sensitive files, 4) Be cautious with MCP servers, 5) Don't open untrusted projects, 6) Review AI-generated code before using it. These steps significantly reduce security risks.

Related Cursor Security Resources

Cursor Security IssuesCursor Privacy ModeCursor MCP VulnerabilitiesCursor SOC 2 CertificationCursor vs Copilot Security

Last updated: January 16, 2026