The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added a critical remote code execution vulnerability in Ray, the open-source distributed computing framework, to its Known Exploited Vulnerabilities (KEV) Catalog after confirming exploitation in the wild.
Tracked as CVE-2025-62593, the flaw affects Ray versions prior to 2.52.0 and carries a CVSS 4.0 score of 9.4. CISA added the issue to the KEV catalog on August 17, 2026, requiring U.S. federal civilian agencies to remediate it by August 20.
The agency’s listing states that the vulnerability could allow remote code execution and specifically warns that developers using Ray as a development tool may be exposed when using Firefox or Safari.
Critical Ray RCE Flaw Exploited
The vulnerability is rooted in Ray’s handling of HTTP API endpoints, including /api/jobs and /api/job_agent/jobs/. An attacker can abuse these interfaces to submit or trigger execution of arbitrary code against a vulnerable local Ray instance.
The weaknesses are categorized as CWE-94, improper control of code generation (commonly known as code injection), and CWE-352, cross-site request forgery (CSRF).
Code-injection weaknesses occur when externally influenced input can alter the intended syntax or behavior of generated code, potentially enabling unauthorized command execution.
According to CISA, Ray’s browser-oriented protection relied on checking whether the User-Agent request header began with “Mozilla.” That safeguard is insufficient because the Fetch specification permits manipulation of the header under certain conditions.
When chained with a DNS rebinding attack, a malicious website or malvertising-delivered page can interact with a locally accessible Ray service through a victim’s Firefox or Safari browser session.
The attack scenario is particularly significant for developers, researchers, and ML engineering teams who run Ray locally while browsing the web.
A successful exploit does not necessarily require the Ray dashboard or API to be publicly exposed to the internet; instead, the attacker can attempt to pivot through the browser to reach a service listening on a local interface or trusted network address.
The practical result could be arbitrary command execution in the context of the user or workload running Ray. Ray Project addressed CVE-2025-62593 in version 2.52.0.
Mitigation
Organizations should immediately identify Ray installations across developer workstations, CI/CD runners, container images, Kubernetes workloads, and cloud-based data-processing clusters, then upgrade all affected deployments to version 2.52.0 or later.
Teams should also verify that dependency lockfiles, base images, and automated build pipelines do not retain vulnerable versions.
CISA advises stakeholders to apply vendor mitigations, assess each asset’s internet exposure, and follow Binding Operational Directive 26-04 risk-based patching guidance.
Where mitigation is unavailable, organizations should discontinue use of the affected product. CISA has not identified known ransomware use of this vulnerability, but its KEV inclusion confirms real-world exploitation and makes rapid remediation essential.
Detect, investigate, and respond faster with in-browser data inspection from ANY.RUN. Gain complete phishing visibility to strengthen your SOC and reduce MTTR
