CISA gives federal agencies just three days to patch actively exploited Ray AI framework flaw
CISA added a critical remote-code-execution flaw in the open-source Ray AI framework to its Known Exploited Vulnerabilities catalog on August 17, 2026, giving federal agencies only three days instead of the usual fourteen to patch. The bug lets attackers hijack a developer's machine via Firefox or Safari using a DNS rebinding attack, then reach into a company's private network. Ray is used by companies including Amazon $AMZN and Apple $AAPL.
CISA added CVE-2025-62593, a critical (CVSS 9.4) remote-code-execution flaw in the open-source Ray AI compute framework, to its Known Exploited Vulnerabilities catalog on August 17, 2026, confirming active exploitation and giving federal civilian agencies until August 20 to patch, a compressed three-day window instead of the standard fourteen. The flaw exists because Ray's dashboard checked only whether a browser's User-Agent header started with 'Mozilla,' a defense that Firefox and Safari allow scripts to bypass; combined with a DNS rebinding attack, this lets an attacker execute code on a developer's machine simply by getting them to visit a malicious website or view a malicious ad. Ray's own security advisory warns the compromised browser can then act as a bridge into a company's private network, reaching Ray clusters that were never meant to be exposed to the internet. Ray is used by major technology companies including Amazon $AMZN and Apple $AAPL, and The Register cited an industry estimate that it runs inside roughly sixty percent of Fortune 500 companies. Unpatched Ray clusters have previously been hijacked for cryptocurrency mining in a campaign researchers call ShadowRay 2.0. The fix is available in Ray version 2.52.0.