CVE-2026-15976
Official description Straight from the sourceThe vendor's or NVD's own wording, published unedited. Authoritative, but often terse — it says what broke, rarely what to do.
NVD · uneditedSGLang 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) fallback enables pickle deserialization of .bin files.
In the news
Third-party coverageSurfaced from public web coverage — external links open in a new tab.
Technical summary Written by usOur analysis, written from the advisory, the CVSS vector and the affected-version data. It adds context the advisory leaves out, and never invents facts that are not in the source.
dbcve analysis · high confidenceSGLang contains a critical RCE vulnerability in the /update_weights_from_disk endpoint which uses torch.load() with weights_only=False, allowing arbitrary pickle deserialization of .bin model weight files. An attacker who can control or supply a malicious .bin file can achieve remote code execution.
Verify against the referenced sources before acting — the references below are authoritative for this CVE, this summary is not.
Affected products & versions What the vendor confirmedThe version ranges the vendor confirmed as vulnerable. If your version sits inside a range here, treat yourself as exposed until you have upgraded.
NVD · CPE data<= 0.5.15CVSS breakdown How the score is builtThe industry scoring standard. It rates how the flaw is reached, what it takes to exploit, and what an attacker gains — the score is derived from those, not the other way round.
From the vector- Attack vector
- Network
- Complexity
- Low
- Privileges
- None
- User interaction
- None
- Scope
- Unchanged
- Confidentiality
- High
- Integrity
- High
- Availability
- High
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Am I affected? How to checkSteps we derive from the advisory and the affected-version data, so you can decide whether this CVE reaches your setup. They are a guide, not a scan — your own configuration is the authority.
dbcve checksWork through these to decide whether this CVE applies to you.
-
Check installed SGLang versionRun `pip show sglang` or `pip list | grep sglang` to identify the installed version numberAffected if The installed version is 0.5.15 or lower (any version <= 0.5.15)
-
Verify /update_weights_from_disk endpoint existsInspect SGLang server source code for the route handler definition at the /update_weights_from_disk path, or check API documentation for available endpointsAffected if The endpoint is present in the codebase (vulnerable code path exists)
-
Confirm torch.load is called with weights_only=FalseSearch the SGLang source code for torch.load() calls within the weight update handler and inspect whether the weights_only parameter is set to False or omittedAffected if torch.load() is invoked without weights_only=True in the weight loading logic
-
Check network exposure of the weight update APIReview server startup configuration and network bindings to determine if the API endpoint is accessible over network (not localhost-only)Affected if The endpoint is exposed to untrusted network sources
You are affected if running SGLang version 0.5.15 or lower AND the /update_weights_from_disk endpoint exists with torch.load() called without weights_only=True.
Generated from the published advisory. Verify against your own configuration.
Remediation Closing itWhat it takes to close this. Where a vendor fix exists we point at it; where none exists we say so plainly, and can build one. Effort estimates are scoped from the advisory, not from your codebase.
dbcve · scopedRestrict network access to the /update_weights_from_disk endpoint and validate/sanitize all model weight files before loading. Consider implementing a secure loading mechanism or switching to weights_only=True if compatible with legitimate use cases.
Latest stable SGLang release (check GitHub security advisories for the specific version number containing the fix)
- Check the official SGLang GitHub repository (github.com/sgl-project/sglang) for security advisories and releases addressing this vulnerability
- Identify the current SGLang version in use (typically via `pip show sglang` or checking environment)
- Upgrade to the latest stable SGLang release that includes the security fix: run `pip install --upgrade sglang`
- Alternatively, if a specific fixed version is mentioned in security advisories, install that version explicitly: `pip install sglang==<fixed_version>`
- After upgrading, verify the fix is applied by checking that `torch.load()` calls in the codebase no longer use `weights_only=False` when loading untrusted model weights
- As a mitigation before patching, avoid loading model weights from untrusted or unverified HuggingFace repositories
- If you must load external weights, ensure they are from trusted sources and have been verified
Generated from the published advisory — verify against the referenced sources before acting.
- Consultation4.0 h
- Implementation8.0 h
- Testing6.0 h
- Review / QA4.0 h
An estimate, not a bill — we confirm scope with you before any work starts. Need it this week? Rush from $6,176.
Scan for this in your stack
Free · runs locallyCheck whether your project pulls in CVE-2026-15976 — or any other known-vulnerable package — straight from your lock files. Free and open source; it runs locally and uploads nothing.
References Go to the primary sourcePrimary sources — vendor advisories, patches and trackers. Where our summary and a reference disagree, the reference wins.
Primary sourcesPractitioner notes
ContributedPeer-ranked notes from engineers who’ve handled CVE-2026-15976 in production — separate from our analysis above.
The advisory tells you what broke. It rarely tells you what actually worked. If you’ve dealt with this one, that detail is what the next engineer is searching for.
- The version that genuinely resolved it — not the one the vendor claimed
- A config change or rule that shut the vector down
- A gotcha in the upgrade path that cost you an afternoon
No notes yet
Be the first to add a field note for this CVE — a mitigation you’ve verified, a version caveat, or a link to a working fix. Sign in above to contribute.
A place for practitioners to share what actually worked: a mitigation you’ve tested, a configuration change, a version- or environment-specific caveat, or a link to a verified patch. The most useful notes rise to the top as peers upvote them, so the signal stays high.
- Verified mitigations, workarounds, and config changes
- Version or environment caveats, and links to real fixes
- No weaponised exploit code, or anything meant to cause harm
- No spam, self-promotion, credentials, or personal data