Over 30 vulnerabilities, collectively called IDEsaster, have been discovered in AI-powered IDEs, enabling data exfiltration and remote code execution through prompt injection and legitimate feature abuse. These security flaws affect popular tools like GitHub Copilot, Cursor, Roo Code, and others, highlighting the need for trusted sources and system hardening. #IDEsaster #PromptInjection
Keypoints
- Over 30 vulnerabilities have been disclosed in AI-powered Integrated Development Environments.
- Attacks leverage prompt injection primitives combined with legitimate features for data exfiltration and RCE.
- Popular IDEs like GitHub Copilot, Cursor, and Roo Code are affected, with some CVEs assigned.
- Attack vectors include bypassing guardrails, auto-approving file writes, and manipulating configuration files.
- Recommendations emphasize using trusted projects, monitoring MCP servers, and applying the principle of least privilege.
Read More: https://thehackernews.com/2025/12/researchers-uncover-30-flaws-in-ai.html