CVE-2026-31942: LibreChat has IDOR in API Keys Management that allows any authenticated user to overwrite other users' API keys

Synopsis

An Insecure Direct Object Reference (IDOR) vulnerability exists in LibreChat versions up to and including 0.7.6, affecting the API keys management endpoint (PUT /api/keys). Any authenticated user can inject a userId parameter into a request body to overwrite another user's stored API key configuration for providers such as OpenAI, Anthropic, and Azure. Successful exploitation lets an attacker silently redirect a victim's AI conversations through attacker-controlled keys or break the victim's service entirely by supplying invalid keys. A fix is available in version 0.8.3-rc1; a patched-image rebuild is available on HarborGuard for environments running an affected version.

HarborGuard Coverage

Exploit Conditions

• Network reachabilityRequired

  The vulnerable endpoint is exposed over the network, so an attacker must be able to reach the LibreChat service via HTTP.

• AuthenticationRequired

  Any low-privilege authenticated account is sufficient; the attacker only needs a valid login to the LibreChat instance.

• Victim interactionNot required

  No action is required from the victim; the attacker submits a crafted PUT request directly without any social engineering.

• Attack complexityDetail

  The exploit is straightforward and condition-free, requiring only the ability to add a userId field to the request body with no race conditions or special environment setup.

Blast Radius

• Overwrites another user's stored API key configuration for providers such as OpenAI, Anthropic, and Azure, silently redirecting that user's AI conversations through attacker-controlled credentials.
• Allows the attacker to monitor or log conversations routed through the substituted API keys on infrastructure they control.
• Denies service to the targeted user by replacing valid API keys with invalid ones, causing all AI requests from that account to fail.