An authorization bypass vulnerability in AutoGPT allows authenticated users to access arbitrary execution results, posing a significant risk to data confidentiality.

The vulnerability exists within the get_graph_execution_results endpoint, where the application fails to verify ownership of the graph_exec_id parameter. This flaw allows any authenticated user to retrieve sensitive execution data that they are not authorized to view.

The ability to access arbitrary execution results through an authorization bypass can lead to the exposure of proprietary AI agent configurations, sensitive data processed by agents, and internal operational workflows. Given the CVSS score of 7.7, this is a high-severity issue that could result in unauthorized information disclosure and loss of intellectual property.

Immediate Action: Upgrade to AutoGPT version 0.6.16 or later to ensure proper authorization checks are enforced on the get_graph_execution_results endpoint.

Proactive Monitoring: Review API access logs for anomalous patterns, specifically looking for repeated calls to the get_graph_execution_results endpoint by users requesting multiple distinct graph_exec_id parameters.

Compensating Controls: Implement strict API gateway controls or network-level access restrictions to limit access to the AutoGPT API to authorized users only, reducing the attack surface until the update can be deployed.

Public Exploit Available: No — exploit_available (unknown).

This vulnerability presents a clear risk to data privacy within the AutoGPT environment. Organizations should prioritize updating their deployments to version 0.6.16 immediately to patch the authorization logic. Failure to remediate this flaw may allow malicious actors with valid credentials to harvest sensitive execution data across the platform.