The Bypass Flaw
Security researchers at Fog Security discovered a critical authorization bypass vulnerability in Amazon Quick, an AWS business intelligence service. The flaw involved the platform’s AI Chat Agents, which are automatically provisioned when Amazon Quick is activated. Because administrators cannot use standard Identity and Access Management policies to govern the chatbot, they must rely on custom permission profiles to restrict access. While the user interface correctly hid the chat feature from blocked users, the backend API did not verify these permissions. By intercepting network traffic and sending direct HTTP requests, a restricted user could successfully interact with the AI agent, bypassing organization-wide blocks.
Impact and Response
The vulnerability created a significant blind spot for enterprise security teams attempting to control shadow AI usage. The platform integrates deeply with corporate data sources such as CRMs and databases, making strict access controls essential. Researchers confirmed the flaw did not allow cross-tenant data exposure but thoroughly compromised intra-account security boundaries. AWS silently patched the flaw without issuing a public advisory or notifying customers, categorizing the risk severity as none. This response has raised concerns about the visibility of security patches for integrated AI services within the AWS ecosystem, as administrators could not confirm whether their environments had been affected.
Source: Cyber Security News
