CVE-2025-54868: LibreChat Information Disclosure Flaw

CVE-2025-54868 Overview

CVE-2025-54868 is an authorization bypass vulnerability in LibreChat, an open-source ChatGPT clone with additional features. The vulnerability exists in versions 0.0.6 through 0.7.7-rc1 where an exposed testing endpoint (/api/search/test) allows unauthorized access to chat data stored in the Meilisearch engine. This flaw enables attackers to read chat conversations from arbitrary users without proper authentication or access control checks.

Critical Impact

Unauthorized attackers can access and read private chat conversations from any user in the system, leading to significant data exposure and privacy violations.

Affected Products

• LibreChat versions 0.0.6 through 0.7.7-rc1
• Self-hosted LibreChat instances with Meilisearch integration enabled
• Deployments exposing the /api/search/test endpoint

Discovery Timeline

• 2025-08-05 - CVE-2025-54868 published to NVD
• 2025-08-26 - Last updated in NVD database

Technical Details for CVE-2025-54868

Vulnerability Analysis

This vulnerability represents a Broken Access Control flaw (CWE-285: Improper Authorization) where a testing endpoint was inadvertently left exposed in production builds. The /api/search/test endpoint was designed for development and debugging purposes to query the Meilisearch search engine directly. However, this endpoint lacked proper access control mechanisms, allowing any network-accessible request to query and retrieve stored chat messages without authentication.

The vulnerability allows attackers to bypass the normal authorization flow that should restrict users to only viewing their own chat conversations. By directly querying the search endpoint, attackers can enumerate and extract chat data belonging to any user in the system.

Root Cause

The root cause of this vulnerability is the inclusion of development and testing routes in production code without adequate access control. The /api/search/test route in api/server/routes/search.js provided direct access to the Message.meiliSearch() function without verifying user authorization. While other routes in the search module were protected by requireJwtAuth middleware, the test endpoint allowed direct queries to be executed against the Meilisearch index containing all user chat messages.

Attack Vector

An attacker can exploit this vulnerability remotely over the network without any authentication. The attack requires no user interaction and can be performed by simply sending HTTP GET requests to the vulnerable endpoint with search query parameters. The attacker can craft search queries to retrieve sensitive chat content, user conversations, and potentially other indexed data from the Meilisearch engine.

// Vulnerable code removed in security patch (api/server/routes/search.js)
// Source: https://github.com/danny-avila/LibreChat/commit/0e8041bcac616949c42a68dfb8f108ccc4db5151

-router.get('/test', async function (req, res) {
-  const { q } = req.query;
-  const messages = (
-    await Message.meiliSearch(q, { attributesToHighlight: ['text'] }, true)
-  ).hits.map((message) => {
-    const { _formatted, ...rest } = message;

The patch removes the vulnerable /test endpoint entirely, along with the /sync endpoint which also lacked proper authorization checks.

Detection Methods for CVE-2025-54868

Indicators of Compromise

• HTTP GET requests to /api/search/test endpoint with query parameters
• Unusual access patterns to search-related API endpoints from unauthenticated sources
• High volume of search queries originating from single IP addresses
• Access logs showing requests to /api/search/test or /api/search/sync endpoints

Detection Strategies

• Monitor web server access logs for requests targeting /api/search/test and /api/search/sync endpoints
• Implement alerting rules for any HTTP requests to deprecated or removed testing endpoints
• Deploy web application firewall (WAF) rules to block access to known vulnerable endpoints
• Review Meilisearch query logs for anomalous search patterns or bulk data retrieval attempts

Monitoring Recommendations

• Enable verbose logging on LibreChat instances to capture all API requests
• Configure SIEM rules to detect reconnaissance activity against search endpoints
• Monitor for data exfiltration patterns indicative of chat data harvesting
• Implement rate limiting on all search-related API endpoints to slow potential exploitation

How to Mitigate CVE-2025-54868

Immediate Actions Required

• Upgrade LibreChat to version 0.7.7 or later immediately
• If immediate upgrade is not possible, manually remove or disable the /api/search/test and /api/search/sync routes
• Review access logs to determine if the vulnerability was exploited before patching
• Notify affected users if unauthorized access to chat data is confirmed

Patch Information

The vulnerability has been addressed in LibreChat version 0.7.7. The fix removes the exposed testing endpoints entirely from the codebase. The security patch is available in commit 0e8041bcac616949c42a68dfb8f108ccc4db5151. For detailed information about the vulnerability and remediation, refer to the GitHub Security Advisory GHSA-p5j8-m4wh-ffmw.

Workarounds

• Block access to /api/search/test and /api/search/sync at the reverse proxy or load balancer level
• Implement network segmentation to restrict access to LibreChat instances from untrusted networks
• Deploy a web application firewall (WAF) with rules to block requests to vulnerable endpoints
• Remove or comment out the vulnerable route handlers in api/server/routes/search.js if patching is delayed

# Nginx configuration to block vulnerable endpoints
location ~* ^/api/search/(test|sync)$ {
    deny all;
    return 403;
}