Which versions of cosmos-cuda are malicious?

The following PyPI versions of cosmos-cuda are flagged malicious: 9999.0.0, 9999.0.1. Treat any of these as compromised.

How do I remediate cosmos-cuda if I installed it?

cosmos-cuda is built to steal secrets, so assume every credential the build or runtime could read is compromised. Remove it from your project and lockfile, then rotate ALL exposed secrets — npm/registry tokens, cloud keys, CI/CD secrets, SSH keys, and any .env values — from a known-clean machine. Audit logs for unauthorized use of those credentials.

I already installed cosmos-cuda — how do I know if it already compromised me?

If cosmos-cuda was installed, its post-install payload may already have run. Removing the package does not undo that — check outbound connections and credential use from the install window onward.

Is cosmos-cuda part of a known attack campaign?

Yes — cosmos-cuda is associated with the GENERIC-standard-pypi-install-pentest, IN-MAL-2026-010570 campaign. Packages in the same campaign often share infrastructure and payloads, so check whether any related packages are also in your dependency tree.

How do I find cosmos-cuda across my projects, lockfiles, and CI pipelines?

Search your lockfiles and build artifacts for cosmos-cuda (2 malicious versions).

Malicious package

cosmos-cudaPyPI

Advisory published Jul 14, 2026Updated Jul 15, 2026

cosmos-cuda is a confirmed malicious PyPI package (MAL-2026-10617) that steals credentials and exfiltrates sensitive data (malicious versions 9999.0.0, 9999.0.1). Do not install it — remove it immediately and rotate any exposed credentials.

Malicious code in cosmos-cuda (PyPI)

MAL-2026-10617

Immediate action

Remove the package, then rotate any secrets the build/runtime could reach.

pip uninstall cosmos-cuda

What this malware does

The sdist-only package cosmos-cuda 9999.0.0 executes a beacon function _beacon('build') from setup.py during sdist build/install and _beacon('import') from cosmos_cuda/init.py at import. Both code paths collect the installer's hostname, current username, and current working directory, then transmit them to the hardcoded external host oob.asyncrun.in via a DNS lookup of a crafted FQDN (socket.getaddrinfo) and via an HTTPS/HTTP GET request (urllib.request.urlopen). The package ships only as an sdist and uses version 9999.0.0 so that pip's highest-version resolution selects it when a name collision exists against an internal package resolved through --extra-index-url, forcing setup.py execution during install and producing arbitrary code execution and exfiltration on installers whose environment resolves the name.

Installing the package or importing the module exfiltrates basic information about the host, and the package has no other purpose.

Category: PROBABLY_PENTEST - Packages looking like typical pentest packages, but also anything that looks like testing, exploring pre-prepared kits, research & co, with clearly low-harm possibilities.

Campaign: GENERIC-standard-pypi-install-pentest

Reasons (based on the campaign):

The package contains code to exfiltrate basic data from the system, like IP or username. It has a limited risk.
The package overrides the install command in setup.py to execute malicious code during installation.

Malicious versions

2 flagged

9999.0.09999.0.1

Indicators of compromise (SHA-256)

bd8651969d2a82aa0c96bae9bce8bffaa3053dc4093fcafc614f7cda24b19542

c1a96e5776ec64356f639738134c95a2f33233a4275d2d0ddd6b174c297c64ee

Detection & response playbook

Credential / info stealer

Find it
Search your lockfiles and build artifacts for cosmos-cuda (2 malicious versions).
If you installed it — respond
cosmos-cuda is built to steal secrets, so assume every credential the build or runtime could read is compromised. Remove it from your project and lockfile, then rotate ALL exposed secrets — npm/registry tokens, cloud keys, CI/CD secrets, SSH keys, and any .env values — from a known-clean machine. Audit logs for unauthorized use of those credentials.
Did it already run?
If cosmos-cuda was installed, its post-install payload may already have run. Removing the package does not undo that — check outbound connections and credential use from the install window onward.

Frequently asked questions

No. cosmos-cuda on PyPI has been identified as a malicious package (versions 9999.0.0, 9999.0.1 flagged). It should be removed immediately — do not install or keep it in your dependency tree.

Campaign

GENERIC-standard-pypi-install-pentestIN-MAL-2026-010570

References

Credits

Amazon Inspector · finder
Kamil Mańkowski (kam193) · reporter

Detect & block this

O3 blocks cosmos-cuda-class packages before install and in CI — and if it already ran, its runtime egress monitoring catches the credential exfiltration and severs the channel.

Malware detection Runtime egress monitoring

Explore

All malware advisories How O3 detects malware Vulnerability database