PatchSiren cyber security CVE debrief
CVE-2026-14890 SGLang CVE debrief
CVE-2026-14890 is a critical vulnerability in SGLang's expert-parallel backup subsystem. The subsystem exposes a ZeroMQ PULL socket on a routable network interface without authentication or deserialization safeguards. This allows an attacker to provide a malicious pickle file, resulting in unauthenticated remote code execution when the feature is enabled and the service is reachable over the network. Organizations using SGLang with the expert-parallel backup feature enabled should prioritize patching. Network administrators and security teams should assess their exposure and implement compensating controls if necessary.
- Vendor
- SGLang
- Product
- Unknown
- CVSS
- CRITICAL 9.1
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-07-16
- Original CVE updated
- 2026-08-10
- Advisory published
- 2026-07-16
- Advisory updated
- 2026-08-10
Who should care
Organizations using SGLang with the expert-parallel backup feature enabled should prioritize patching. Network administrators and security teams should assess their exposure and implement compensating controls if necessary. SGLang users with the expert-parallel backup feature enabled are at risk of unauthenticated remote code execution.
Technical summary
The vulnerability exists in SGLang's expert-parallel backup subsystem, which exposes a ZeroMQ PULL socket without authentication or deserialization safeguards. An attacker can exploit this by providing a malicious pickle file, leading to unauthenticated remote code execution when the feature is enabled and the service is network-reachable. The expert-parallel backup feature in SGLang is exposed on a routable network interface without authentication or deserialization safeguards.
Defensive priority
High priority should be given to patching SGLang installations with the expert-parallel backup feature enabled. In the absence of a patch, consider disabling the feature or restricting network access to the service.
Recommended defensive actions
- Apply the vendor's official patch as soon as available.
- Disable the expert-parallel backup feature if not in use.
- Restrict network access to the ZeroMQ PULL socket.
- Monitor for suspicious activity related to SGLang services.
- Inventory SGLang installations and assess exposure.
- Review compensating controls for exposed systems while remediation is scheduled and verified.
- Track exceptions, retest remediated assets, and close the item only after evidence is documented.
Evidence notes
The CVE record was published on 2026-07-16T16:19:00.413Z and last modified on 2026-07-16T19:16:44.403Z. The NVD entry is currently Awaiting Analysis. References include CERT and GitHub links. The expert-parallel backup feature in SGLang is exposed on a routable network interface without authentication or deserialization safeguards, allowing an attacker to provide a malicious pickle file. This results in unauthenticated remote code execution when the feature is enabled and the service is reachable over the network. The vulnerability exists in SGLang's expert-parallel backup subsystem. The CVE record was last modified on 2026-07-16T19:16:44.403Z.
Sources and references
Verified primary and authoritative sources
-
CVE-2026-14890 CVE Program record
Publisher, destination, and source semantics verified
URL: https://www.cve.org/CVERecord?id=CVE-2026-14890
CVE Program - Official CVE Program record with source-provided CVE metadata.
-
CVE-2026-14890 NVD vulnerability detail
Publisher, destination, and source semantics verified
URL: https://nvd.nist.gov/vuln/detail/CVE-2026-14890
NIST National Vulnerability Database - Official NIST NVD detail page and source-specific vulnerability assessment.
Supplemental references
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/blob/main/python/sglang/srt/elastic_ep/expert_backup_manager.py
-
Source reference
Unverified legacy reference
URL: https://www.kb.cert.org/vuls/id/326070
af854a3a-2127-422b-91ae-364da2661108
Methodology and review provenance
AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.