Executive brief
The Strands Agents Tools service includes an http_request tool used by AI agents to make HTTP requests on behalf of users. Due to incorrect authorization logic, this tool fails to properly validate whether a user or agent has permission to make specific HTTP requests, potentially allowing unauthorized data access or manipulation of external systems that the tool can reach.
Technical details
The vulnerability is an authorization bypass in the http_request tool within Strands Agents Tools. The tool fails to properly enforce access controls when processing HTTP requests, allowing callers to make requests that should be restricted based on their privilege level or intended scope. An authenticated agent or user with network connectivity to the http_request tool can exploit this flaw to access restricted resources or perform unauthorized operations. The vulnerability does not require special preconditions beyond basic agent execution access. No patch information is available in the advisory.
Affected products
- Strands Agents Tools <UNKNOWN>
Timeline
- 2026-09-22: disclosed