Amazon Quick Flaw Exposed Restricted AI Chat Agents to Unauthorized Users

Fog Security discovered that Amazon QuickSight’s AI Chat Agent enforced custom permissions only in the UI while the backend chat-agent API lacked server-side authorization, enabling basic-account users to send POST requests to the API and receive AI-generated responses despite being restricted. The flaw (CWE-862) could expose insights from internal datasets and undermine governance and compliance; Fog Security disclosed via HackerOne and AWS rolled out a fix within about a week, with the issue limited to the same AWS account and no cross-account data leakage reported.