CVE records

Every CVE as the CVE Program publishes it, refreshed every ten minutes, with NVD's analysis, CISA's known-exploited status and a healthcare tag. Subscribers get the CVEs that match their own systems on My program, by email as they are published. Where the data comes from: sources.

Updated under an hour ago (2026-10-19 08:00 UTC)

What this means for you

This week: 5,437 CVEs published, 355 critical, and 7 newly confirmed exploited. Add your stack to see whether any of them hit you.

Do this next

Start free and add your stack

Free account, 1 product

This week's critical CVEs

Exploited in the last 30 days

Published in the last 7 days: 5,437; Show in the list
Critical, last 7 days: 355; Show in the list
Healthcare-tagged, last 7 days: 4; Show in the list
Known exploited (CISA KEV): 1,747; Show in the list

Showing CVEs that name the product lmsys:sglang. Show all products

15 CVEs, showing 1 to 15

CVE	Severity	Title	Published
CVE-2026-15978	High 7.5	SGLang contains a model weight exfiltration vulnerability when no API keys are configured, as SGLang will expose two endpoints that allow a remote attacker to trigger distributed weight broadcasting…	Jul 30, 2026
CVE-2026-15977	High 7.5	SGLang contains a credential leakage vulnerability in the /server_info endpoint, which will return API keys and SSL keyfile information when only the --admin-api-key is configured.	Jul 30, 2026
CVE-2026-15976	Critical 9.8	SGLang contains a RCE vulnerability when attempting to load model weights from a HuggingFace repository, specifically within the /update_weights_from_disk, where torch.load(..., weights_only=False) f…	Jul 30, 2026
CVE-2026-15974	Medium 6.5	SGLang contains an SSRF and local file read in the multimodal generation endpoint /v1/chat/completions due to unsanitized image_url, allowing access to internal metadata, secrets, and services.	Jul 30, 2026
CVE-2026-15971	Critical 9.8	SGLang contains an RCE vulnerability when the optional dumper subsystem is enabled, allowing for a sandbox escape when DUMPER_SERVER_PORT is set, enabling code execution on inference requests.	Jul 30, 2026
CVE-2026-15969	Critical 9.8	SGLang contains an unauthenticated RCE in /load_lora_adapter_from_tensors via bypass of SafeUnpickler’s incomplete denylist, allowing arbitrary command execution through crafted base64-encoded pickle…	Jul 30, 2026
CVE-2026-14890	Critical 9.1	SGLang uses an expert-parallel backup subsystem that exposes a ZeroMQ PULL socket on a routable network interface that does not contain authentication or deserialization safeguards, allowing an attac…	Jul 16, 2026
CVE-2026-10775	Medium 5.3	A vulnerability was determined in sgl-project SGLang up to 0.5.11.	Jun 3, 2026
CVE-2026-7304	Critical 9.8	SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads() will…	May 18, 2026
CVE-2026-7302	Critical 9.1	SGLangs multimodal generation runtime is vulnerable to an unauthenticated path traversal vulnerability, allowing an attacker to write arbitrary files anywhere the server process has write access, by…	May 18, 2026
CVE-2026-7301	Critical 9.8	SGLangs multimodal generation runtime scheduler's ROUTER socket binds to 0.0.0.0 by default and contains a sink that calls pickle.loads() on incoming messages, enabling RCE when exposed to the intern…	May 18, 2026
CVE-2026-5760	Critical 9.8	SGLang's reranking endpoint (/v1/rerank) achieves Remote Code Execution (RCE) when a model file containing a malcious tokenizer.chat_template is loaded, as the Jinja2 chat templates are rendered usin…	Apr 20, 2026
CVE-2026-3989	High 7.8	SGLangs `replay_request_dump.py` contains an insecure pickle.load() without validation and proper deserialization.	Mar 12, 2026
CVE-2026-3060	Critical 9.8	SGLang' encoder parallel disaggregation system is vulnerable to unauthenticated remote code execution through the disaggregation module, which deserializes untrusted data using pickle.loads() without…	Mar 12, 2026
CVE-2026-3059	Critical 9.8	SGLang's multimodal generation module is vulnerable to unauthenticated remote code execution through the ZMQ broker, which deserializes untrusted data using pickle.loads() without authentication.	Mar 12, 2026

CVSS is the assigning CNA's score where it gives one, otherwise NVD's. The healthcare tag is a documented heuristic (a list of clinical and health IT vendors, and clinical terms in the description), not a classification. The CVE record is authoritative.