Skip to main content

Fastchat EUVDEUVD-2026-23780

| CVE-2026-6608 MEDIUM
Always-Incorrect Control Flow Implementation (CWE-670)
2026-04-20 VulDB GHSA-f3q6-69f3-vwch
5.5
CVSS 4.0 · NVD
Share

Severity by source

NVD PRIMARY
5.5 MEDIUM
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:L/VA:N/SC:N/SI:N/SA:N/E:P/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X

Primary rating from NVD · only source for this CVE.

CVSS VectorNVD

Attack Vector
Network
Attack Complexity
Low
Privileges Required
None
User Interaction
None
Scope
X

Lifecycle Timeline

6
PoC Detected
Apr 22, 2026 - 20:22 vuln.today
Public exploit code
Analysis Generated
Apr 20, 2026 - 06:28 vuln.today
CVSS changed
Apr 20, 2026 - 06:22 NVD
5.3 (MEDIUM) 5.5 (MEDIUM)
EUVD ID Assigned
Apr 20, 2026 - 06:15 euvd
EUVD-2026-23780
Analysis Generated
Apr 20, 2026 - 06:15 vuln.today
CVE Published
Apr 20, 2026 - 05:15 nvd
MEDIUM 5.5

Blast Radius

ecosystem impact
† from your stack dependencies † transitive graph · vuln.today resolves 4-path depth
  • 30 pypi packages depend on fschat (25 direct, 5 indirect)

Ecosystem-wide dependent count for version 0.2.36.

DescriptionCVE.org

A vulnerability was detected in lm-sys fastchat up to 0.2.36. Impacted is the function add_text of the component Arena Side-by-Side View Handler. The manipulation results in incorrect control flow. The attack can be launched remotely. The exploit is now public and may be used. The root cause was fixed in commit 34eca62 for gradio_block_arena_named.py, but three other files were missed.

AnalysisAI

Information disclosure in lm-sys FastChat up to version 0.2.36 allows remote unauthenticated attackers to manipulate the add_text function in the Arena Side-by-Side View Handler, resulting in incorrect control flow that exposes sensitive data. The vulnerability has publicly available exploit code and affects the web-based arena comparison interface. A partial fix was applied in commit 34eca62 to gradio_block_arena_named.py, but three additional affected files remain unpatched.

Technical ContextAI

The vulnerability resides in FastChat's Arena Side-by-Side View Handler, a component responsible for comparing multiple language models side-by-side during inference. The add_text function processes user input without proper validation or sanitization, leading to CWE-670 (Improper Control Flow Management). The affected component is implemented across multiple Python files that interact with the Gradio framework (evidenced by gradio_block_arena_named.py). The root cause involves incorrect control flow logic that fails to properly enforce data isolation between competing models or sessions, allowing attackers to redirect execution paths and leak information. The partial upstream fix in commit 34eca62 demonstrates awareness of the issue, but the incomplete remediation across related handler files leaves multiple exploitation vectors open.

RemediationAI

Immediate action: Apply the upstream fix from commit 34eca62 (https://github.com/lm-sys/FastChat/issues/3834) to gradio_block_arena_named.py, then identify and patch the three remaining affected files mentioned in the CVE description by cross-referencing the FastChat repository issue tracker. Since no official patched version release is confirmed, manually backport commit 34eca62 and related fixes to your deployed version. Temporary workaround while awaiting complete patches: Disable the Arena Side-by-Side View Handler component or restrict network access to the arena evaluation endpoints via firewall rules, limiting exposure to trusted internal networks only. Note that disabling arena functionality may impact model comparison workflows, so evaluate this trade-off against your organization's evaluation requirements. Monitor the GitHub repository for a confirmed full release that addresses all four affected files. Test any patches thoroughly in a non-production environment before deployment, as the incomplete upstream fix suggests potential interactions between the partially-patched files.

Share

EUVD-2026-23780 vulnerability details – vuln.today

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy