Nvidia Patches Critical Security Vulnerabilities in AI Server Software
Technology giant Nvidia has released a security update for its Triton Inference Server, addressing three critical vulnerabilities identified by cybersecurity firm Wiz.
The flaws, designated CVE-2025-23319, CVE-2025-23320, and CVE-2025-23334, could potentially allow attackers to gain full control of AI servers, steal sensitive data, or manipulate model responses, according to Wiz.
“Wiz Research found a chain of vulnerabilities that, when combined, could let an attacker with no prior access take full control of an AI server,” Nir Ohfeld, head of vulnerability research at Wiz, told Cointelegraph.
“The attack typically starts with a minor bug that causes the server to leak a small piece of secret internal data. An attacker can then use that data to trick one of the server’s legitimate features into giving them control over a private system component,” Ohfeld explained. “This initial foothold allows for privilege escalation leading to complete server takeover.”
Triton is Nvidia’s open-source software designed to optimize the deployment and execution of AI models on servers.
While the total number of Triton users remains unspecified, notable enterprises using the software include Microsoft, Amazon, Oracle, Siemens, and American Express. Nvidia reported over 25,000 companies utilized its AI stack in 2021.
An Nvidia spokesperson declined comment beyond referencing the company’s security advisory and recommending the patch.
“The single most important step is to update to the patched version of the Nvidia Triton Inference Server (version 25.07 or newer),” Ohfeld advised. “This directly fixes the entire vulnerability chain.”
As of now, there is no reported evidence these specific vulnerabilities have been actively exploited in the wild. Nvidia Triton serves as a widely adopted platform for AI workloads.
Related: 5 smart contract vulnerabilities: How to identify and mitigate them
Security vulnerabilities hamper emerging technologies
Security vulnerabilities continue to challenge emerging technologies in 2025, notably in the cryptocurrency sphere.
According to blockchain security auditor Hacken, access flaws and smart contract bugs have resulted in $3.1 billion in crypto asset theft during the first half of 2025, surpassing total losses from the previous year.
Experts warn that AI agents and quantum computing may introduce further cyber threats in the coming landscape.
