CriticalPyPI3 affected versions

xinference

xinference has 3 versions on PyPI flagged in 1 incident we track. Other versions of this package are not implicated - only the exact versions listed below were named by the source advisories.

Affected versions

Check what you actually have installed with pip show xinference.

xinference@2.6.2
xinference@2.6.1
xinference@2.6.0

The incident that flagged it

Critical22 Apr 2026

Xinference PyPI package backdoored on import

JFrog flagged three consecutive xinference releases (2.6.0-2.6.2) on PyPI carrying a TeamPCP-style payload embedded in xinference/__init__.py that fires on import. The malware base64-encodes a stage-1 wrapper containing the comment # hacked by teampcp, then spawns a detached subprocess that bundles harvested data into love.tar.gz and POSTs it with a custom X-QT-SR: 14 header. TeamPCP publicly denied involvement, claiming a copycat. 600,000+ downloads of malicious wheels.

Versions named here: 2.6.0, 2.6.1, 2.6.2

What to do if you installed one of these

Treat the machine as compromised rather than simply upgrading. Packages in these incidents typically execute at install time, so the payload has already run with whatever credentials were in the environment.

1.Rotate every credential that machine or CI job could reach - registry tokens, cloud keys, SSH keys, CI secrets - from a different machine.
2.Pin to a version not listed above, then reinstall from a clean lockfile.
3.Clear caches and private mirrors. Artifactory, Nexus, Verdaccio, and devpi routinely keep serving a tarball after the public registry has pulled it.
4.Read the incident write-up above - recommended actions differ per incident, and some have specific indicators of compromise worth grepping your logs for.

Check your whole project

xinference is one of many packages in these incidents, and it is usually pulled in as a transitive dependency rather than one you installed directly. Paste your requirements.txt to check every dependency at once - it runs in your browser and nothing is uploaded.

Check my requirements.txt All flagged PyPI packages

Version data comes from the public advisories cited on each incident page. We list only versions those advisories name - we don't infer additional ones. Think this is wrong or incomplete? Tell us.