🔍 Read the full analysis: Microsoft’s AI Agent Framework Was Hijacked On PyPI To Steal Credentials. Who Must Respond? on ThorstenMeyerAI.com
Get movie nights delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Two versions of Microsoft’s official Python package agent-framework, 1.26.1 and 1.26.2, carried credential-stealing code and were briefly available on PyPI on November 17, 2026. PyPI quarantined both releases, but researchers say the malicious code could run when Python started, even if the framework was not imported. The number of affected installations is not known; users who installed either version are advised to rebuild environments and rotate accessible credentials.
Two releases of Microsoft’s official Agent Framework Python package on PyPI contained credential-stealing code, according to security advisories and researchers who analyzed the files. Versions 1.26.1 and 1.26.2 were published on November 17 and then quarantined by PyPI; developers and organizations that installed either version are advised to treat the environment as exposed and rotate credentials it could access.
The affected package was agent-framework, the PyPI package linked from Microsoft’s installation instructions. The GitHub Advisory Database says versions 1.26.1 and 1.26.2 included a credential stealer absent from the last clean release, 1.26.0, and rates the incident critical. The malicious code was added alongside the working framework, so a successful installation or normal operation would not necessarily alert a user.
According to the published timeline, version 1.26.1 appeared at 20:35 UTC on November 17 and PyPI quarantined it at 21:06 UTC. The attacker then uploaded 1.26.2 at 21:08 UTC; PyPI quarantined that release at 21:16 UTC, according to StepSecurity. Both versions were removed from availability. The source says PyPI’s current version is 1.26.0, released November 12.
The advisory says the added startup hook could run when Python started in an affected environment, without the user importing agent_framework. Researchers said the code sought credentials and tokens available to developer machines and build environments. The advisory’s response guidance is to rebuild affected environments from a clean base using version 1.26.0 and rotate every credential the environment could reach, including CI and GitHub Actions secrets.
agent-framework on PyPI · UTC
Two poisoned releases, one evening
The official package of Microsoft’s Agent Framework, step by step. Blue: what happened. Green: the clean version.
November 12 to 18, 2026
- Clean 1.26.0 publishedStill the current version on PyPI
- Malicious 1.26.1 appearsFramework works normally, a credential stealer runs whenever Python starts
- PyPI quarantines 1.26.1After about 31 minutes
- Attacker uploads 1.26.2Renamed files, also targets GitHub Actions tokens
- PyPI quarantines 1.26.2Eight minutes later
- Malware databases list itMAL-2026-20101 (OpenSSF), GHSA-764m-3ppc-5m73 (GitHub)
Sources: GitHub Advisory Database GHSA-764m-3ppc-5m73, Nov 18, 2026; OSV MAL-2026-20101, Nov 17, 2026; PyPI project data, checked Nov 19, 2026; StepSecurity analysis (quarantine times).
Graphic: Powered by Thorsten Meyer AI
Why Build Pipelines Face Exposure
The incident reaches beyond people deliberately testing an AI framework. A developer laptop, continuous-integration runner, build server, or container image could have installed a compromised release through a direct install or an automatic update. Since the code could run at Python startup while the framework itself continued to work, ordinary testing might not reveal the compromise.
Build environments can hold credentials for cloud accounts, code repositories, package registries, and AI services. If such credentials were accessible to the malicious code, the risk could extend from one package installation to systems and services those credentials can access. That is a potential consequence, not confirmation that any particular account was used or breached. Researchers cited in the source found no remote-code-loading second stage in the code they inspected, and described the malware as not persisting beyond the installed package. Those findings do not establish whether credentials were stolen or misused.
The second upload also highlights a response challenge: removing one malicious version does not by itself prove that the publishing route is secure. StepSecurity said the first quarantine did not immediately revoke the attacker’s publishing access; JFrog similarly noted that quarantining a release does not revoke a compromised publishing credential. Publisher access and package removal are separate issues, and defenders need to check both when responding to a supply-chain incident.
Popular right now

KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream

Dyson Cyclone V10 Origin Cordless Vacuum

Bose QuietComfort Wireless Noise Cancelling Headphones – Twilight Blue Limited Edition
As an affiliate, we earn on qualifying purchases.
How the Two Releases Appeared
Microsoft’s Agent Framework is an open-source toolkit for building AI agents in Python and coordinating them in workflows. The PyPI package agent-framework serves as an umbrella package that installs the framework’s core code and integrations. The source says Microsoft released the clean 1.26.0 version on November 12, before the two suspect releases appeared five days later.
The package was not described as an unrelated lookalike. The source says Microsoft’s own installation instructions directed users to the official PyPI package, and security company Wiz described it as receiving almost 200,000 downloads a month. That figure is a reported monthly download count; it does not show how many downloads occurred during the brief incident window or how many installations were affected.
Several security organizations analyzed the releases, including Socket, StepSecurity, Wiz, JFrog, and Endor Labs. The OpenSSF-run Open Source Malware database listed the incident as MAL-2026-20101 on November 17, while the GitHub Advisory Database published advisory GHSA-764m-3ppc-5m73 on November 18. These reports establish that the packages were analyzed and flagged; they do not provide a confirmed count of victims.
“The malicious code runs without importing agent_framework.”
— GitHub Advisory Database
Victim Count and Credential Use
The number of affected users and systems has not been published in the source material. Wiz’s monthly download estimate does not establish how many people installed either malicious version, and the brief availability window does not by itself determine the number of exposed environments. It is also not clear whether any collected credentials were used to access accounts or systems.
The source reports no Microsoft security advisory at the time of publication and provides no detailed account of how the publishing access was compromised, who was responsible, or whether credentials were exfiltrated from specific victims. Researchers’ findings about the inspected malware describe its behavior; they are not evidence that every listed type of credential was present or stolen in a given environment.
Rebuild, Rotate, and Monitor
Organizations should check package-install records, lockfiles, build logs, and container histories for versions 1.26.1 and 1.26.2. The advisory recommends treating any environment that installed either release as exposed, rebuilding it from a clean base on 1.26.0, and rotating every credential the environment could access. Security teams should also review relevant account and cloud activity for suspicious use; the source does not report a confirmed impact assessment or a count of affected installations.
Microsoft’s stated position was that it had made some repositories private while investigating malicious content. The source reports that researchers notified Microsoft and that PyPI had quarantined the releases, but does not give a date for a fuller public update. Further information about the publishing-access compromise, any affected credentials, and confirmed downstream activity remains dependent on investigation and disclosures from Microsoft, PyPI, and affected organizations.
Key Questions
Which package versions were affected?
The affected PyPI releases were agent-framework 1.26.1 and 1.26.2. The source identifies 1.26.0, released November 12, as the last clean release.
Did someone have to import the framework for the code to run?
No. The GitHub Advisory Database says the malicious startup hook could run when Python started in an affected environment, without importing agent_framework.
What should someone do if they installed one of the releases?
The advisory recommends treating the environment as exposed, rebuilding from a clean base on version 1.26.0, and rotating every credential it could reach, including CI and GitHub Actions secrets.
How many users were affected?
No confirmed victim or installation count is provided in the source material. Wiz’s estimate of almost 200,000 downloads a month is not a count of downloads or affected users during the incident window.
Has Microsoft explained how the malicious releases were published?
The source reports that Microsoft told The Register it had made some repositories private while investigating malicious content. It does not report a detailed explanation of how publishing access was obtained or a full Microsoft security advisory.
Source: ThorstenMeyerAI.com
