The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Microsoft's new NLWeb AI protocol, designed to enable agentic AI across websites, contained a critical vulnerability allowing unauthorized access to sensitive files and API keys. Discovered by security researchers, the flaw was patched in July 2025, but its existence highlights significant security risks in rapidly deployed AI systems.[AI generated]
Why's our monitor labelling this an incident or hazard?
The event involves an AI system (NLWeb) and a security vulnerability that allowed unauthorized access to sensitive AI-related data (API keys for LLMs). This directly relates to the use and development of an AI system and has led to a significant security risk that could cause harm, including financial loss and malicious cloning of AI agents. Since the flaw existed and was exploitable before the fix, this constitutes an AI Incident due to the realized risk and potential harm caused by the AI system's vulnerability.[AI generated]