What happened
CVE-2026-111704 (CVSS 9.8, CWE-306) verified via NVD REST API (published 2026-10-16 by VulnCheck): LangChat's REST controllers are reachable unauthenticated because no Sa-Token authentication interceptor is registered, letting remote attackers read plaintext user passwords, retrieve LLM provider API keys, create super-admin accounts, and modify/delete data.
Why it matters
A total unauthenticated compromise of an LLM chat/agent gateway: full disclosure of every user's credentials and every upstream LLM provider API key. An attacker can burn the victim's OpenAI/Anthropic/etc. quota, impersonate any user, and read all AI conversations. This is the model-credential exposure pattern that matters most to AI deployments.
Attack vector
No Sa-Token interceptor is registered on REST controllers; unauthenticated requests to /api/v1/auth/users return plaintext passwords, /api/v1/auth/* allows creating super-admin accounts, and stored LLM provider API keys are retrievable.
Affected systems
LangChat ≤ commit 2a6285a (github.com/LangChat/langchat)
Mitigation
Upgrade beyond the affected commit; ensure an authentication interceptor covers all REST controllers and rotate any exposed LLM provider keys/credentials. See VulnCheck advisory and the AigcUserController reference.