Critical vulnerability in Ray framework allows remote code execution

A critical vulnerability in the open-source Ray framework, used for scaling Python and machine-learning workloads, is being actively exploited by attackers, according to the Cybersecurity and Infrastructure Security Agency (CISA). The flaw, tracked as CVE-2025-62593, allows for remote code execution on vulnerable Ray systems, according to a recent report by The Register.

The vulnerability, rated 9.4 under CVSS v4, was disclosed in November 2025 and allows attackers to exploit browsers like Firefox and Safari to achieve remote code execution (RCE) on a vulnerable Ray system. This is possible because vulnerable Ray versions attempt to block browser requests by checking if the User-Agent header starts with "Mozilla," a check that Firefox and Safari can bypass. Developers running Ray could trigger the exploit by visiting a malicious website or encountering a compromised ad, enabling attackers to use DNS rebinding to access the local Ray service.

The vulnerability impacts developers running development and testing environments, potentially allowing arbitrary shell code execution on their machines. It can also be used to attack network-adjacent Ray instances. Ray 2.52.0 addresses the flaw, and CISA has mandated a three-day remediation window for US federal agencies due to the perceived risk.

Source: The Register