CVE-2026-5632

7.3

assafelovic · GPT Researcher

GPT Researcher versions 3.4.3 and earlier contain a critical vulnerability where all HTTP REST API endpoints and WebSocket interfaces are exposed without any authentication or authorization.

Executive summary

The GPT Researcher application is vulnerable to a complete authentication bypass, allowing any unauthenticated remote attacker to gain full control over the system, including file manipulation and LLM API usage.

Vulnerability

The application architecture completely lacks authentication middleware or access control checks for its FastAPI-based REST API and WebSocket endpoints. This allows any unauthenticated network user to interact with critical functions, including file uploads, file deletions, and the triggering of research tasks.

Business impact

The lack of authentication poses a severe risk to organizational security, as it allows unauthorized parties to exfiltrate sensitive research reports, delete critical data, and consume expensive LLM API credits. Given the exposure of file management functions, an attacker could potentially achieve remote code execution or facilitate further system compromise. While the CVSS score varies by reporting source, the total lack of access control and the ability to trigger automated actions remotely elevate this to a critical priority for all deployments.

Remediation

Immediate Action: Since no official patch is currently available, administrators must restrict network access to the application immediately by placing it behind a secure reverse proxy that enforces robust authentication, such as Nginx with Basic Auth or OIDC.

Proactive Monitoring: Monitor server access logs for anomalous request patterns, specifically looking for high volumes of API calls, unexpected file deletion attempts, or requests from unknown source IP addresses.

Compensating Controls: Ensure the application is not exposed to the public internet, utilize network-level access control lists (ACLs) to limit access to trusted internal IP ranges, and run the service with the least-privileged user account possible to mitigate the impact of potential file system manipulation.

Exploitation status

Public Exploit Available: Yes, a published proof-of-concept exists and is documented in the referenced GitHub issue report.

Analyst recommendation

This vulnerability represents a critical security failure that exposes the entire application to unauthorized control. Given the absence of a vendor-provided patch, organizations must treat this as an urgent priority and implement immediate network-level isolation or reverse proxy authentication to prevent unauthorized access. Do not expose this service to the public internet under any circumstances until a formal authentication framework is integrated into the software.

Sources

Originally found and disclosed by Yu-Bao (VulDB User), with VulDB CNA Team (coordinator), per the CVE Program record.