Check Point Discloses Two High-Impact Vulnerabilities in LiteLLM AI Gateway Allowing Low-Privilege Users to Steal API Keys
Check Point has disclosed two vulnerabilities in LiteLLM, an open-source AI gateway downloaded more than 90 million times per month. The issues, tracked as CVE-2026-35029 (CVSS 8.7) and CVE-2026-59823 (CVSS 5.3), affect all versions before 1.83.9 and allow low-privilege users to take over the gateway, read server files, and steal API keys stored for providers such as OpenAI, Anthropic, and Azure OpenAI.
A recent incident at an internet company illustrated the risk when an intern successfully called the GPT-4 API using a corporate key. Investigation revealed that the LiteLLM instance lacked proper authorization checks on its configuration endpoint, enabling any authenticated account to alter settings arbitrarily.
What LiteLLM Does
LiteLLM acts as a central proxy that unifies access to more than 140 AI vendors. It handles routing, load balancing, key management, usage monitoring, and cost control so developers do not need to manage individual provider APIs directly. Because it stores real API credentials for enterprise use, compromise of the gateway can lead to unauthorized consumption of AI resources and exposure of business logic derived from call patterns.
CVE-2026-35029: Unauthorized Configuration Changes
The first flaw resides in the /config/update endpoint. The code performs no role verification, so any authenticated user—including interns or accounts obtained through phishing—can modify runtime settings. Attackers can read arbitrary server files containing stored keys, redirect requests to attacker-controlled servers, hijack administrator accounts, or implant malicious code for remote code execution.
CVE-2026-59823: SSRF Vector
The second issue is a server-side request forgery vulnerability. When combined with the configuration flaw, it allows an attacker to force the gateway to reach internal databases, metadata services, or other sensitive systems. In cloud environments this often serves as the initial step toward obtaining temporary credentials and lateral movement.
Scope and Remediation
LiteLLM 1.83.9 contains fixes for both vulnerabilities. Because the component is frequently embedded inside larger AI platforms rather than deployed standalone, many organizations may be unaware of its presence. Security teams should:
- Upgrade via pip install litellm --upgrade
- Restrict access to the management interface so only users with the proxy_admin role can call configuration endpoints
- Rotate all stored API keys if unauthorized access cannot be ruled out
- Search codebases and dependency files for references to litellm
- Enable detailed logging to detect anomalous configuration changes
The disclosure underscores that core security controls remain essential even in rapidly adopted AI infrastructure.
Related articles
Top 5 Cybersecurity Events of the Week: Barracuda Root Exploits, cPGuard Supply-Chain Attack, APT Flaw, Xen Escape Bugs, and OpenSSH Authentication Bypass
This week’s Jet CSIRT roundup highlights five critical security issues affecting widely deployed infrastructure. Two high-severity vulnerabilities in Barracuda Email Security Gateway and other appliances allow unauthenticated attackers to obtain root access via authorization bypass and command injection. A supply-chain compromise of cPGuard versions 5.88.12–5.88.19 enabled theft of password hashes, database files, and SSH keys from at least 32 customer servers before the malicious update was removed. A flaw in the APT package manager (CVE-2026-109518) lets network attackers substitute malicious packages that execute with root privileges on Debian and Ubuntu systems. Two Xen hypervisor bugs (CVE-2026-102703 and CVE-2026-102704) permit a compromised guest VM to read and write memory belonging to dom0 and other guests. Finally, an OpenSSH authentication bypass on musl-based Alpine Linux distributions allows passwordless login to key-only accounts when long password hashes trigger a crypt() return value of “*”.
CISA and FBI Warn of Global Cyber Attacks on Industrial Refrigeration and Cold Chain Controllers
CISA, the FBI, and multiple U.S. agencies issued a joint advisory on November 5 detailing malicious cyber activity targeting industrial refrigeration systems and cold chain monitoring equipment worldwide. Attackers are exploiting unpatched or internet-exposed temperature controllers in supermarkets and cold storage facilities through authentication bypass, arbitrary code injection, and weak configurations. Once inside the control panels, threat actors force units into continuous heating and defrost cycles, the opposite of their intended function, while also causing denial of visibility that prevents operators from seeing real equipment status. The advisory highlights risks including reduced system lifespan, fire hazards, stock loss, and potential consumer harm from spoiled products stored outside safe temperature ranges. Sectors most affected include public health, chemical manufacturing, food and agriculture, and refrigerated transportation. Recommended mitigations focus on network segmentation, removal of remote access from the public internet, strong authentication, firmware updates, and regular offline backups.
Critical Flaw in React Server Components Enables Unauthenticated Remote Code Execution
The React team disclosed security advisory GHSA-xqc2-xf4c-x3h3 describing a high-impact vulnerability in React Server Components that can lead to remote code execution. The flaw affects packages react-server-dom-webpack, react-server-dom-parcel, and react-server-dom-turbopack across multiple version ranges from 19.0.0 through 19.3.0. Exploitation requires no authentication, privileges, or user interaction, allowing a remote attacker to execute code on the server by sending a specially crafted request containing multiple typed arrays alongside attacker-controlled data. No CVE identifier was assigned, and the advisory notes that the issue is distinct from the previously reported React2Shell vulnerability tracked as CVE-2025-55182. Affected applications are limited to those running React code on the server with supported frameworks and bundlers; client-only React usage is unaffected. Patches are available in versions 19.0.9, 19.1.10, 19.2.9, and 19.3.1, with immediate updates strongly recommended. The React team also linked similar typed-array handling issues in Next.js Cache Components, suggesting a broader class of errors.
Plex Media Server Vulnerabilities Exploited in Wild via Plex2Shell Chain
Plex has disclosed active exploitation of four vulnerabilities in Plex Media Server versions 1.43.2 and earlier. The flaws, tracked as CVE-2026-96652, CVE-2026-96654, CVE-2026-96655 and CVE-2026-96656, allow unauthenticated remote code execution when chained together. Attackers are leveraging server-side request forgery and token-handling weaknesses to write arbitrary files on exposed servers. Proof-of-concept code named Plex2Shell was published on 16 October, after which CISA added the issues to its Known Exploited Vulnerabilities catalog. Compromised systems have been observed deploying cryptocurrency miners and proxy relays. The software is widely used both as a standalone media server and as an add-on for QNAP and other NAS devices, increasing the potential attack surface.