Critical Security Vulnerability Found in Ray Open Source Framework
CISA gives federal agencies 3 days to patch actively exploited flaw.
1Critical Ray Vulnerability Discovered
CISA has identified a critical security vulnerability in the Ray open source framework used for scaling Python and machine learning workloads. Tracked as CVE-2025-62593, this flaw received a CVSS v4 score of 9.4. The vulnerability allows remote code execution on vulnerable Ray systems through Firefox or Safari browsers. Ray is widely used and supported by major tech companies including Amazon, Apple, and OpenAI.
2Attack Mechanism
Vulnerable Ray versions attempt to identify and block browser requests by checking for 'Mozilla' in the User-Agent header. However, Firefox and Safari permit scripts using the Fetch API to modify this header. A developer can trigger the exploit simply by visiting a malicious website or receiving a harmful ad in an affected browser. Attackers can then use DNS rebinding to access the local Ray service and execute arbitrary code.
3Affected Environments
This vulnerability primarily impacts development and testing environments using Ray. Project developers warn that users falling victim to phishing attacks or malicious ads can be exploited, allowing execution of arbitrary shell code on their machines. The attack can also leverage browsers as confused deputies to target Ray instances running within private corporate networks.
4CISA's Urgent Patch Directive
CISA has mandated US federal civilian agencies to remediate this vulnerability within 3 days instead of the standard 14-day window. While the agency didn't explain the urgency, it marked the 'known ransomware use' field as unknown. Binding Operational Directive 26-04 empowers CISA to impose such short remediation timelines for especially risky vulnerabilities.
5Ray's Growth and Adoption
Ray is an open source framework that helps scale Python and ML workloads from local to cluster environments with minimal code changes. Now managed by PyTorch Foundation, the project originated at UC Berkeley and was commercialized by Anyscale in 2019. As of October 2025, Ray boasts over 237 million total downloads with 7 million weekly downloads, showing nearly tenfold year-over-year growth.
6Enterprise Usage Statistics
Product analysis site NextSprints estimates Ray has 1 million monthly active users and is deployed by 60% of Fortune 500 companies. Security advisories attribute the vulnerability to Ray's historical lack of authentication on critical endpoints. Ray's security model traditionally assumed clusters would run within trusted, isolated networks, delegating authentication to surrounding infrastructure.
7Solution in Ray 2.52.0
Ray 2.52.0 addresses the flaw and introduces optional token-based authentication as additional protection against unauthorized access. However, this feature remains disabled by default. The project continues recommending cluster deployment within controlled networks rather than treating authentication as a substitute for proper isolation. This update is particularly crucial for users running development environments.
8Industry Impact
This vulnerability poses significant risks for developers in machine learning and AI fields. Given Ray's widespread adoption and rapid growth, prompt patching is critical. CISA's unusually short remediation window underscores the flaw's potential severity. Immediate action by organizations is vital to prevent potential large-scale cyber attacks that could exploit this vulnerability across numerous development environments.