---
id: "CVE-2026-44484"
page: "https://infrastack.ai/live/CVE-2026-44484"
kind: "vulnerability"
aliases: ["CVE-2026-44484", "GHSA-w37p-236h-pfx3"]
title: "Compromise of PyTorch Lightning PyPi Package Versions"
severity: "critical"
cvss: 9.8
epss: 0.00392
exploited: false
status: "open"
category: "package"
products: [{"name": "pytorch-lightning", "fixed": [], "affected": "= 2.6.2", "ecosystem": "pypi"}, {"name": "pytorch-lightning", "fixed": [], "affected": "= 2.6.3", "ecosystem": "pypi"}, {"name": "lightningai pytorch lightning", "fixed": [], "affected": "= 2.6.2", "ecosystem": "vendor"}, {"name": "lightningai pytorch lightning", "fixed": [], "affected": "= 2.6.3", "ecosystem": "vendor"}, {"via": "cpe", "name": "lightning", "fixed": [], "affected": "= 2.6.2", "ecosystem": "pypi"}, {"via": "cpe", "name": "lightning", "fixed": [], "affected": "= 2.6.3", "ecosystem": "pypi"}]
ecosystems: ["pypi", "vendor"]
tags: ["ci", "cwe-506", "cwe-829", "package", "pypi", "python", "pytorch-lightning"]
not_affected_if: ["pinned-before-2.6.2"]
published_at: "2026-05-07T00:52:55+00:00"
first_seen_at: "2026-11-16T16:13:27+00:00"
updated_at: "2026-11-16T16:38:06+00:00"
sources: [{"source": "ghsa", "url": "https://github.com/advisories/GHSA-w37p-236h-pfx3", "first_seen_at": "2026-11-16T16:13:27+00:00"}, {"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-44484", "first_seen_at": "2026-11-16T16:38:06+00:00"}]
exploit_sources: [{"type": "epss", "score": 0.00392, "url": "https://api.first.org/data/v1/epss?cve=CVE-2026-44484"}]
---

# Compromise of PyTorch Lightning PyPi Package Versions

Link for people: https://infrastack.ai/live/CVE-2026-44484 (this .md is for agents)

Versions 2.6.2 and 2.6.3 of the pytorch-lightning package on PyPI were published with malicious code that the maintainer describes as consistent with credential harvesting. An attacker who gets this code installed can take API keys, tokens, SSH keys and service account credentials from the machine.

_Summary, check and fix written by a model from the sources; labelled as such._

## Affected

- pypi:pytorch-lightning affected = 2.6.2 fixed no fixed version listed
- pypi:pytorch-lightning affected = 2.6.3 fixed no fixed version listed
- vendor:lightningai pytorch lightning affected = 2.6.2 fixed no fixed version listed
- vendor:lightningai pytorch lightning affected = 2.6.3 fixed no fixed version listed
- pypi:lightning affected = 2.6.2 fixed no fixed version listed
- pypi:lightning affected = 2.6.3 fixed no fixed version listed

## How to check

- Run `pip show pytorch-lightning` and look for version `2.6.2` or `2.6.3`
- Check lockfiles and requirements files for `pytorch-lightning==2.6.2` or `pytorch-lightning==2.6.3`
- Review build and CI logs for installs of those versions and for unexpected outbound connections from those jobs

## Fix

- pytorch-lightning: move to `2.6.1`, which the maintainer recommends; the malicious versions were quarantined from PyPI
- Rotate all credentials and secrets that may have been exposed, per the maintainer's advisory

## Timeline

- 2026-11-16T16:13:27+00:00 ghsa https://github.com/advisories/GHSA-w37p-236h-pfx3
- 2026-11-16T16:38:06+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-44484

## Documents

- [advisory] Maintainer's advisory: api.github.com https://api.github.com/repos/Lightning-AI/pytorch-lightning/security-advisories/GHSA-w37p-236h-pfx3
- [advisory] Maintainer's advisory: github.com https://github.com/Lightning-AI/pytorch-lightning/security/advisories/GHSA-w37p-236h-pfx3

## From the sources

# Security Advisory: Compromise of PyTorch Lightning PyPI Package Versions

**Published:** 2026-04-30  
**Last Updated:** 2026-05-12 

**Github Advisory:** [CVE-2026-44484](https://github.com/advisories/GHSA-w37p-236h-pfx3)

We have identified a security incident affecting certain versions of one of our PyPI packages.

## What happened

We have determined that one or more released versions of this package have been compromised and include malicious code.

Our current investigation indicates that the affected versions have introduced functionality consistent with a credential harvesting mechanism. We are still actively analysing the scope and behaviour of the code.

At this stage, the root cause of the compromise is still under investigation.

## What versions are affected

We are currently working to confirm the exact set of impacted versions.

We have determined the following versions as affected and ask that you delete them from your systems:

- `2.6.2`
- `2.6.3`

We will update this advisory if the versions impacted by this vulnerability change.

## What you should do immediately

If you have installed or are running any potentially affected versions:

- Assume the environment may be compromised  
- Immediately rotate all credentials and secrets that may have been exposed, including:  
  - API keys  
  - Access tokens  
  - SSH keys  
  - Service account credentials  
- Rebuild affected systems from a known clean state  
- Pin PyTorch Lightning to version `2.6.1`  
- Review logs for any suspicious or unauthorised activity  

## Actions we have taken

- Quarantined malicious versions from PyPI  
- Recommend using version `2.6.1`: https://github.com/Lightning-AI/pytorch-lightning/releases/tag/2.6.1  
- Revoked and rotated all internal credentials associated with our release process  
- Initiated a full investigation into the compromise  

## Ongoing investigation

We are actively working to:

- Identify the exact mechanism of compromise  
- Confirm the full set of affected versions  
- Determine the behaviour and impact of the malicious code  
- Assess any downstream impact to users  

We will provide updates as soon as more information becomes available.

## Commitment to transparency

We take the security of our users and the integrity of our software supply chain extremely seriously.

We will continue to share timely and accurate updates as our investigation progresses.

## Contact

If you have questions or believe you may be impacted, please contact us at:

**security@lightning.ai**

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_
