Microsoft's AI Web Project Faces Security Flaws
Originally Published 5 months ago — by The Verge

Researchers discovered a critical security flaw in Microsoft's new NLWeb protocol, which allows remote reading of sensitive files, including API keys, due to a path traversal vulnerability. Microsoft patched the issue but has not issued a CVE, raising concerns about security oversight in AI-related protocols. The flaw could have severe consequences for AI agents relying on exposed API keys, emphasizing the need for careful security practices in deploying new AI features.
