LangChat Critical Flaw Lets Attackers Steal Keys and Create Admin Accounts
A missing authentication check in LangChat allows unauthenticated attackers to access REST controllers, steal API keys, and create super-admin accounts.
Key points
- CVE-2026-111704 is rated Critical with a CVSS score of 9.3
- The flaw affects LangChat versions up to commit 2a6285a
- Attackers can read plaintext passwords and retrieve LLM provider keys
The National Vulnerability Database has published CVE-2026-111704, identifying a critical security flaw in the LangChat application. This vulnerability allows unauthenticated remote attackers to access protected REST controllers due to a missing authentication mechanism. The issue exists in LangChat through commit 2a6285a and is classified as CWE-306.
According to the NVD record, the vulnerability stems from the fact that no Sa-Token interceptor is registered within the application. This oversight exposes sensitive endpoints to anyone on the network. Attackers can query the /api/v1/auth/users endpoint without credentials. This access enables them to read plaintext passwords stored in the system.
Context and Background
The flaw grants attackers significant control over the LangChat environment. Beyond reading user credentials, adversaries can retrieve API keys associated with Large Language Model providers. These keys are critical for connecting to external AI services and represent a high-value target for theft. The exposure of these keys could lead to unauthorized usage or data leakage.
Attackers can also create new super-admin accounts within the affected system. This privilege escalation allows them to maintain persistent access even if initial credentials are changed. Furthermore, the vulnerability permits the modification or deletion of existing data. This broad range of capabilities underscores the severity of the missing authentication control.
Who Is Affected
The vulnerability affects any deployment of LangChat that has not been updated beyond commit 2a6285a. Organizations using this version for internal AI chat functionalities are at risk. The flaw is present in the core REST controller handling, meaning all users of this specific codebase are exposed. There is no indication that specific configurations mitigate the risk.
The NVD record does not specify which downstream products or distributions include this vulnerable code. However, any system running the identified commit range is susceptible. Security teams should verify their current LangChat version against the affected commit hash. The lack of a registered Sa-Token interceptor is the root cause across all affected instances.
What should you do?
Security professionals should immediately audit their LangChat installations to determine if they are running the vulnerable version. Check the commit history or version string against 2a6285a. If the system is affected, isolate it from untrusted networks to prevent exploitation. Monitor logs for unauthorized access to the /api/v1/auth/users endpoint.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: LangChat
- Verify your LangChat version and compare it against the affected commit 2a6285a to determine exposure
- Monitor server logs for unusual activity on the /api/v1/auth/users endpoint from unauthenticated sources
- Rotate any exposed LLM provider API keys and user passwords immediately if compromise is suspected
- Implement network-level restrictions to limit access to the LangChat REST controllers until a fix is available
Step-by-step guide: Risk-Based Vulnerability Management: Prioritize Fixes by Actual Danger
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is the CVSS score for CVE-2026-111704?
The vulnerability has a CVSS score of 9.3, which is rated as Critical by the National Vulnerability Database.
What specific action allows attackers to exploit this flaw?
Attackers can query the /api/v1/auth/users endpoint because no Sa-Token interceptor is registered to block unauthenticated requests.
Can attackers create new admin accounts using this vulnerability?
Yes, the flaw allows unauthenticated remote attackers to create super-admin accounts within the affected LangChat system.
