Remote Code Execution Flaw in Giskard AI Library
CVE-2026-34172

7.7HIGH

Key Information:

Vendor

Giskard-ai

Status
Giskard-oss
Vendor
CVE Published:
31 March 2026

What is CVE-2026-34172?

The Giskard AI library, used for testing and evaluating agentic systems, has a vulnerability that allows for remote code execution. In versions prior to 0.3.4 and 1.0.2b1, the ChatWorkflow.chat(message) method inappropriately processes user input as a Jinja2 template without sandboxing. This oversight can be exploited by developers who pass unsanitized user inputs, potentially leading to class traversal attacks within the Jinja2 environment. As a result, attackers can execute arbitrary code on affected systems. The vulnerabilities have been rectified in the subsequent releases.

Affected Version(s)

giskard-oss < 0.3.4 < 0.3.4

giskard-oss >= 1.0.1a1, < 1.0.2b1 < 1.0.1a1, 1.0.2b1

References

https://github.com/Giskard-AI/giskard-oss/security/adviso...
x_refsource_CONFIRM

CVSS V4

Score:
7.7
Severity:
HIGH
Confidentiality:
High
Integrity:
High
Availability:
High
Attack Vector:
Network
Attack Complexity:
Low
Attack Required:
Physical
Privileges Required:
Undefined
User Interaction:
None

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.