CVE-2026-31228 Trusted-AI CVE debrief

Who should care

Organizations using the Adversarial Robustness Toolbox (ART) version 1.20.1 or earlier should prioritize updating to a secure version. This vulnerability affects users of ART's Kubeflow component, particularly those in machine learning and AI environments. Given the critical severity and high CVSS score, immediate attention is required to prevent potential exploitation.

Technical summary

The Adversarial Robustness Toolbox (ART) contains a remote code execution vulnerability in its Kubeflow component. Specifically, the robustness evaluation function for PyTorch models uses the eval() function without proper sanitization to evaluate user-supplied strings for LossFn and Optimizer parameters. This allows an attacker to provide malicious input that, when evaluated, can execute arbitrary Python code. The vulnerability is tracked as CVE-2026-31228 and has a CVSS score of 9.8, indicating critical severity. The vulnerability exists in ART versions up to 1.20.1.

Defensive priority

High priority should be given to updating ART to a version that addresses this vulnerability. Organizations should review their use of ART's Kubeflow component and ensure that any user-supplied input is properly sanitized.

Recommended defensive actions

• Update ART to a version that addresses the CVE-2026-31228 vulnerability.
• Review and restrict use of the eval() function in ART's Kubeflow component.
• Implement proper input validation and sanitization for user-supplied strings.
• Monitor for and restrict malicious input to ART's robustness evaluation function.
• Consider compensating controls, such as network restrictions or additional monitoring, until a patch is applied.

Evidence notes

The CVE-2026-31228 vulnerability was publicly disclosed on May 12, 2026, and has since been modified on June 30, 2026. The vulnerability affects ART versions up to 1.20.1. The CVSS score of 9.8 indicates critical severity. Limited information is available on known affected scope or vendor remediation workflow.

Sources and references

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.