Researchers bypass browser security in AI coding assistants

# Researchers bypass browser security in AI coding assistants

https://gcdradar.vercel.app/incidents/researchers-bypass-browser-security-in-ai-coding-assistants · AIR-2026-1640 · closed 2026-11-12

- Lane: Research
- Severity: 1 (Research / proof of concept): Research finding demonstrating local file access and code execution without wild exploitation confirmation.
- Fix status: Fix status unknown
- First reported: 2026-10-29
- AI role: Attack on AI
- Products: Cursor, GitHub Copilot, Codex, VS Code, Claude Desktop

## Summary

Researchers discovered vulnerabilities in built-in browsers across five AI coding assistants, enabling code execution and local file access. The flaws affected tools from Cursor, GitHub Copilot, OpenAI, Microsoft, and Anthropic.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Affected and fixed versions

- Affected: Cursor
- Affected: GitHub Copilot
- Affected: Codex
- Affected: VS Code
- Affected: Claude Desktop

## Latest updates

- 2026-10-29: Researchers broke the browsers built into VS Code, Cursor, Copilot, Codex and Claude Desktop (https://helpnetsecurity.com/2026/10/29/research-ai-assistant-browser-vulnerabilities)

## Sources

- 2026-10-29 Help Net Security: Researchers broke the browsers built into VS Code, Cursor, Copilot, Codex and Claude Desktop https://helpnetsecurity.com/2026/10/29/research-ai-assistant-browser-vulnerabilities

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.