External risk intelligence

Improper Certificate Validation in req_llm WebSocket Client Exposes OpenAI Sessions.

CVE advisorySeverity: CRITICAL (CVSS 9.1)

CVE-2026-107685

The vulnerability exists in a library component (req_llm) used by developers to build applications. It is a client-side library dependency, not an internet-facing service, appliance, or gateway itself. Exposure depends entirely on how a developer integrates the library within their own custom application code.

Halo Surface Signal: 1 out of 5 — much less likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

A vulnerability has been identified in a software component used for handling WebSocket connections, specifically affecting the streaming client for OpenAI. This issue allows an attacker positioned between the client and server to potentially intercept and modify data, including sensitive authorization tokens, within these sessions. While not all communication paths are impacted, those relying on this specific client for OpenAI sessions are at risk.

Compromised client validation allows data interception.
Protects sensitive communication channels.
Confirm relevance and assess potential exposure.

Attack Path

How an attacker could exploit the issue

An attacker with network access could compromise WebSocket sessions used for OpenAI interactions by exploiting an improper certificate validation flaw. This vulnerability in the WebSocket streaming client allows an on-path attacker to intercept and alter these sessions, including sensitive authorization tokens.

Network access required.
Unverified TLS certificate triggers vulnerability.
Sensitive tokens can be read or modified.

Live Threat

Current exploitation, exposure, and threat context

This vulnerability could allow an on-path attacker to intercept and alter data transmitted through specific WebSocket streaming connections. When supported, this could affect sensitive information sent over OpenAI WebSocket sessions, such as authorization tokens.

API keys and access tokens at risk.
Interception via untrusted TLS certificates.
Session data could be read or modified.

Operational Fix

Recommended remediation, mitigation, and detection steps

This vulnerability impacts applications using the `req_llm` library's WebSocket streaming client, potentially exposing sensitive tokens. The primary responsibility likely falls on the application development teams who integrate this library, in coordination with infrastructure or platform teams managing the underlying services. The first practical step is to identify all applications that utilize the affected `req_llm` components, determine their exposure to on-path attackers, and confirm ownership for remediation planning.

Application development teams own the issue.
Verify affected applications and their exposure.
Plan remediation based on business risk.

Supplementary metadata

CVSS vector

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Free EASM Trial See How Signal Works

Frequently asked questions

What is the req_llm library?

req_llm is a library for the Elixir programming language designed to help developers interact with Large Language Models. Specifically, it provides tools for streaming data to and from services like OpenAI. It serves as a connector in custom software, managing the complex WebSocket communication required for real-time AI responses.

What does CVE-2026-107685 mean for security?

This CVE involves Improper Certificate Validation, categorized as CWE-295. In plain terms, the software fails to verify the identity of the server it connects to over a WebSocket. Because it skips this vital security check, an attacker can present a fake security certificate to impersonate the legitimate server, allowing them to intercept or tamper with the session.

How can an attacker trigger this vulnerability?

An attacker must be positioned on the network path between your application and the AI service. If they can intercept your traffic, they can present an untrusted TLS certificate which the library will incorrectly accept. Note that this flaw is specific to the WebSocket client; HTTP streaming connections using Finch or HTTP/2 sessions using Mint already perform proper validation and remain secure.

Do I need to worry about this vulnerability?

Halo Surface Signal indicates this is unlikely to be an immediate internet-facing threat because the issue exists inside a developer library, not an exposed service. Relevance depends on whether your custom applications use the affected library to manage OpenAI sessions. You should check if your software relies on req_llm to handle sensitive API keys or OAuth tokens over WebSockets.

How do I respond to this security issue?

Start by auditing your codebase to identify if and where the req_llm library is integrated. Focus on the WebSocket streaming client specifically. Once identified, your development team should verify if you are using a vulnerable version of the library and plan to update to a patched release. Since this is a library dependency, updates must be managed within your application's build process.

References

Written by Nicholas Merritt

Nicholas Merritt is VP of Security Solutions at Halo Security. With more than 25 years across application security, vulnerability prioritization, and offensive security, he helps teams translate threat data into practical exposure validation and remediation focus. He authors Halo Threat Intelligence CVE advisories using Halo Surface Signal and H/A/L/O context to prioritize internet-facing risk.