BentoML versions before 1.4.38 are vulnerable to remote code execution during cloud deployment due to improper input sanitization in shell command construction.

The vulnerability is an OS Command Injection (CWE-78) located in the cloud deployment script. The application fails to sanitize the system_packages parameter, which is interpolated into a shell command and subsequently executed on the BentoCloud build infrastructure.

Successful exploitation allows an attacker to achieve remote code execution on the target CI/CD build tier. This could lead to a full compromise of the deployment pipeline, unauthorized access to sensitive AI models, or the injection of malicious code into downstream production environments. Given the CVSS score of 7.8, this represents a high-severity risk to the integrity and security of the software supply chain.

Immediate Action: Update the BentoML library to version 1.4.38 or later immediately to incorporate the necessary input sanitization fixes.

Proactive Monitoring: Review CI/CD pipeline logs for unexpected shell command executions or anomalous setup.sh modifications during the build process.

Compensating Controls: Restrict access to cloud deployment configurations and enforce strict input validation policies for any parameters utilized in automated deployment scripts.

Public Exploit Available: No (Exploit available: false/unknown)

This vulnerability presents a significant risk to development operations and environment integrity. Organizations utilizing BentoML for cloud-based model deployment must prioritize upgrading to version 1.4.38 to eliminate the command injection vector. Immediate action is required to secure the CI/CD pipeline against potential exploitation.