SGLang allows unauthenticated pickle deserialization through /update_weights_from_tensor when no auth keys are configured, and the SafeUnpickler… (CVE-2026-86793)
SGLang contains a critical vulnerability (CVE-2026-86793) that allows unauthenticated remote code execution via pickle deserialization through the /update_weights_from_tensor endpoint when no authentication keys are configured. The SafeUnpickler protection can be bypassed due to resolvable builtins.import and builtins.getattr, enabling exploitation via pickle REDUCE. This vulnerability has a CVSS score of 9.8, indicating a severe risk of full system compromise if exploited.
AI Analysis
Technical Summary
The vulnerability in SGLang arises from unauthenticated pickle deserialization through the /update_weights_from_tensor endpoint when authentication keys are not set. The SafeUnpickler mechanism intended to restrict unsafe deserialization can be bypassed because builtins.import and builtins.getattr remain accessible, allowing attackers to execute arbitrary code via the pickle REDUCE opcode. This leads to remote code execution without any privileges or user interaction required.
Potential Impact
Successful exploitation allows an unauthenticated attacker to execute arbitrary code on the affected system remotely, resulting in full confidentiality, integrity, and availability compromise. This can lead to complete system takeover, data theft, or service disruption.
Mitigation Recommendations
Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, it is critical to configure authentication keys to prevent unauthenticated access to the /update_weights_from_tensor endpoint. Avoid exposing this endpoint to untrusted networks and monitor for suspicious activity related to pickle deserialization attempts.
SGLang allows unauthenticated pickle deserialization through /update_weights_from_tensor when no auth keys are configured, and the SafeUnpickler… (CVE-2026-86793)
Description
SGLang contains a critical vulnerability (CVE-2026-86793) that allows unauthenticated remote code execution via pickle deserialization through the /update_weights_from_tensor endpoint when no authentication keys are configured. The SafeUnpickler protection can be bypassed due to resolvable builtins.import and builtins.getattr, enabling exploitation via pickle REDUCE. This vulnerability has a CVSS score of 9.8, indicating a severe risk of full system compromise if exploited.
CVSS v3.1
Score 9.8critical
Weaknesses
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
The vulnerability in SGLang arises from unauthenticated pickle deserialization through the /update_weights_from_tensor endpoint when authentication keys are not set. The SafeUnpickler mechanism intended to restrict unsafe deserialization can be bypassed because builtins.import and builtins.getattr remain accessible, allowing attackers to execute arbitrary code via the pickle REDUCE opcode. This leads to remote code execution without any privileges or user interaction required.
Potential Impact
Successful exploitation allows an unauthenticated attacker to execute arbitrary code on the affected system remotely, resulting in full confidentiality, integrity, and availability compromise. This can lead to complete system takeover, data theft, or service disruption.
Mitigation Recommendations
Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, it is critical to configure authentication keys to prevent unauthenticated access to the /update_weights_from_tensor endpoint. Avoid exposing this endpoint to untrusted networks and monitor for suspicious activity related to pickle deserialization attempts.
Technical Details
- Gcve Source
- db.gcve.eu
- Osv Id
- GHSA-m675-99qc-9233
- Osv Schema Version
- 1.4.0
- Aliases
- ["CVE-2026-86793"]
- Database Specific Severity
- CRITICAL
- Cvss Version
- 3.1
Threat ID: 6aa8a1d055bf5e2cf5f3e945
Added to database: 09/15/2026, 01:39:28 UTC
Last enriched: 09/15/2026, 02:03:48 UTC
Last updated: 09/15/2026, 06:25:46 UTC
Views: 7
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.