The kb-ai PyPI package (versions 0.1.0 and 0.1.1) includes code that, when imported, triggers a background thread collecting installer host identifiers such as hostname, platform, Python version, and timestamp. This data is sent via an unconditional POST request to a hardcoded Azure-hosted callback URL. Errors during this process are silently ignored. The package's own comments indicate this is a supply-chain proof-of-concept illustrating a dependency confusion attack. Additionally, the package overrides the setup.py install command to execute malicious code during installation. The package's main functionality is a stub, with the primary purpose being demonstration of this attack vector.

The package exfiltrates basic system information without user consent, which may lead to privacy violations and potential reconnaissance for further attacks. The unconditional network call and silent error handling increase the risk of unnoticed data leakage. The override of the install command to execute malicious code during installation could allow arbitrary code execution on the host system. However, the package is described as a proof-of-concept with limited risk and no known exploits in the wild.

No official patch or remediation is available. Users should avoid installing or importing the kb-ai package versions 0.1.0 and 0.1.1. Since this is a proof-of-concept package demonstrating a supply-chain attack, the best mitigation is to verify package authenticity and source before installation. Monitor for updates from the package maintainer or PyPI for any official fixes or removals.