AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Microsoft’s official Python package for Agent Framework briefly included a credential stealer in versions 1.26.1 and 1.26.2 on November 17, according to security advisories. Teams that installed either version should treat the environment as exposed, rebuild from a clean version, and rotate credentials it could access; the full scope of exposure is unknown.

Two releases of Microsoft’s official Python package for Agent Framework contained credential-stealing code and were briefly available on PyPI on November 17, according to the GitHub Advisory Database and security researchers. Developers and organizations that installed versions 1.26.1 or 1.26.2 should treat affected environments as exposed, rebuild them from a clean base, and rotate credentials those systems could reach.

The affected package is agent-framework, the PyPI package linked from Microsoft’s installation instructions for its open-source toolkit for building and coordinating AI agents. The GitHub Advisory Database says versions 1.26.1 and 1.26.2 included a credential stealer absent from the last clean release, version 1.26.0. The advisory rates the incident critical.

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

PyPI records show version 1.26.1 was published at 20:35 UTC on November 17. StepSecurity says PyPI quarantined it at 21:06 UTC, after about 31 minutes. A second release, 1.26.2, appeared at 21:08 UTC and was quarantined at 21:16 UTC, according to StepSecurity. The package’s current PyPI version was 1.26.0 when checked November 19.

The advisory says the malicious code ran when Python started in an affected environment, even if the user did not import the agent-framework package. The framework continued to work, which could make the compromise less apparent. The advisory recommends treating systems that installed either release as exposed, rebuilding them from a clean base, and rotating every credential they could access, including CI and GitHub Actions secrets.

At a glance
updateWhen: Malicious releases published November 1…
The developmentTwo compromised versions of Microsoft’s agent-framework package were published to PyPI on November 17 and later quarantined, prompting warnings for users to rebuild affected environments and rotate accessible credentials.
Bar chart titled 50 stored requests: who sent them? Based on Wiz's analysis of the requests stored by the attacker's free collection service: 29 from sandboxes, vendor scans and tests; 17 carrying real victim data; 4 from the attacker's own machines. Highlighted: behind the 17 requests, 3 victim cases, which Wiz counts as unique, successful thefts of victim data. Note: only what the free service stored, at most 50 requests; no install count has been published
What the attacker's collection service stored, per Wiz Research (researchers' claim, updated November 18, 2026). Chart: Thorsten Meyer AI.

Credentials at Risk Across Build Systems

The incident could affect more than a developer’s local machine. If a compromised package ran in a CI pipeline, build server, laptop, or container, the malware could reach credentials available to that environment. Those credentials may grant access to cloud accounts, source code, deployment systems, or AI services.

Because the package kept working and the code ran at Python startup, a successful install might not have produced an obvious failure. Teams may need to investigate package and build records to identify affected systems, then rotate credentials and review relevant access. The reports do not establish that every credential reachable from an affected system was stolen or misused.

The second upload also shows why removing a release from PyPI is only one part of containment. StepSecurity and JFrog say quarantining the first version did not immediately revoke the publishing access used to upload the next one. Quarantine can stop new installs from PyPI, but copies already present in caches, mirrors, or container images may remain.

Products Worth Considering

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

As an affiliate, we earn on qualifying purchases.

Two Releases, Minutes Apart

Microsoft’s Agent Framework is an open-source toolkit for building AI agents in Python. The agent-framework package acts as an umbrella package that brings in the framework’s core code and integrations. Version 1.26.0, published November 12, was the last clean release identified in the advisory.

Security researchers including Socket, StepSecurity, Wiz, JFrog, and Endor Labs reported analyzing the malicious releases. Endor Labs noted that the releases it examined from 1.21.0 through 1.26.0 had included source code, while 1.26.1 was published as a wheel only, according to the GitHub advisory. A wheel is a prebuilt Python package file.

The OpenSSF malware database listed the incident as MAL-2026-20101 on November 17. The GitHub Advisory Database published advisory GHSA-764m-3ppc-5m73 early on November 18. Wiz said it reported its findings to Microsoft on November 17. These reports document the package releases and response timeline; they do not establish how the publishing credentials were compromised.

“The credential stealer is not present in the last clean release, 1.26.0.”

— GitHub Advisory Database

Exposure Scope Still Unknown

Available advisories and researcher timelines do not say how many users installed either release, which credentials were accessed, or whether those credentials were used after exposure. They also do not explain how the attacker obtained publishing access or establish whether any downstream systems were affected.

Microsoft’s reported statement says some repositories were made private while the company investigates, but the source material identifies no Microsoft security advisory with further details. PyPI quarantine removes the releases from ordinary installation through the registry; it does not confirm that cached copies, mirrors, or previously built images were removed. Teams will need to check their own dependency and build records.

Teams Should Rebuild and Rotate

Organizations should check dependency lockfiles, build logs, package caches, and container histories for versions 1.26.1 and 1.26.2. If either version was installed, the advisory’s guidance is to rebuild the environment from a clean base using version 1.26.0 and rotate all credentials that environment could reach, including CI and GitHub Actions secrets.

Teams should also review relevant credential use and access records for activity during and after the exposure window, following their security procedures. Microsoft’s investigation and any further advisories may clarify the publishing compromise and impact. The source material does not give a date for additional findings.

Key Questions

Which package versions were affected?

The GitHub Advisory Database identifies agent-framework 1.26.1 and 1.26.2 as malicious. It identifies 1.26.0 as the last clean release.

Who needs to take action?

Any team or individual whose laptop, build server, CI pipeline, or container image installed either affected version should treat that environment as exposed and follow the advisory’s remediation guidance.

What should an affected team do?

Rebuild the affected environment from a clean base on version 1.26.0 and rotate every credential the environment could access, including CI and GitHub Actions secrets.

Does PyPI quarantine mean existing copies are gone?

No. Quarantine prevents ordinary installation of those releases from PyPI, but the source material says copies may remain in caches, mirrors, or images. Teams should check their own systems.

Is it known whether credentials were stolen or misused?

The source material does not establish how many users were affected, which credentials were accessed, or whether any were later misused. Those points remain under investigation or are otherwise unreported.

Source: Thorsten Meyer AI

You May Also Like

Victims Blast Grok For Allegedly Using Their Media In AI Deepfake Training

Survivors allege xAI’s Grok chatbot trained on their images and videos without consent, raising legal and ethical concerns about data sourcing and re-victimization.

Iphone Safety Check: Cut off Access Fast

Keep your iPhone secure by quickly cutting off unauthorized access—discover essential safety steps to protect your device today.

Location Services: What to Leave On—And Off

I can help you understand which location services to keep on or off for optimal privacy and battery life—discover the best practices inside.

How CERT-In Will Test Cybersecurity Vulnerabilities With Claude Mythos

A Moneycontrol headline reports CERT-In has access to Anthropic’s Claude Mythos for vulnerability testing, but scope and results are not disclosed.