Technology newsroom
Critical Vulnerability in Bifrost AI Gateway! Hackers Can Run Commands on Servers Without Authentication 😱
A critical vulnerability (CVSS 9.8) has been found in all Bifrost AI Gateway versions prior to 2.1.0, allowing hackers to remotely execute commands on servers without authentication. Users are advised to update immediately.
📌 Key Takeaways:
- A critical vulnerability, CVE-2026-90898, has been discovered in Bifrost AI Gateway with a severity score of 9.8
- Remote attackers can execute arbitrary code or commands on the server (RCE) without requiring a password or authentication
- Affects all Bifrost versions prior to 2.1.0, and an urgent update is recommended to mitigate the risk
Urgent alert for administrators and developers using Bifrost, an open-source AI Gateway that acts as an intermediary for sending requests to over 20 large language model (LLM) providers. A severe security vulnerability has been discovered.
The vulnerability, tracked under CVE-2026-90898, has been assessed as Critical with a CVSS score as high as 9.8 out of 10, clearly reflecting its danger. This flaw allows malicious actors to send harmful commands to Bifrost-installed servers remotely, and most concerningly, attackers do not require any authentication credentials.
🚨 Attack Mechanism and Impact
The attack can be successfully executed with just a single, specially crafted HTTP request. Once the Bifrost server receives such a request, it immediately processes the embedded command, allowing hackers to gain complete control over the server. This could lead to theft of sensitive data, malware installation, or using the server as a base to attack other systems. This vulnerability affects all Bifrost versions prior to 2.1.0 where the Management Authentication feature is enabled.
🛡️ Protection and Resolution
The Bifrost developers have already released a patch in version 2.1.0. Therefore, all administrators using Bifrost are strongly advised to check their version and update to the latest version as soon as possible to close this attack vector and prevent potential damage to organizational systems and data.
💬 Does your organization use an AI Gateway to connect to LLM services, and what measures do you have in place to check for vulnerabilities in these tools?