🔍 Read the full analysis: Who Can Act On The PyPI Hijack Of Microsoft’s AI Agent Framework? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Two versions of Microsoft’s Python Agent Framework package, 1.26.1 and 1.26.2, were briefly available on PyPI on November 17, 2026, with credential-stealing code, according to security advisories. PyPI quarantined both releases, but anyone who installed either should treat the environment as exposed, rebuild from a clean base and rotate credentials it could access.
Two releases of Microsoft’s official Python Agent Framework package on PyPI contained credential-stealing code and were available for a short period on November 17, according to the GitHub Advisory Database and security researchers. PyPI quarantined both versions, but developers and organizations that installed 1.26.1 or 1.26.2 should treat those environments as exposed and rotate any credentials they could reach.
The affected package was agent-framework, the PyPI package referenced by Microsoft’s installation instructions. The GitHub Advisory Database says versions 1.26.1 and 1.26.2 contained a credential stealer absent from the preceding clean release, 1.26.0, and rates the incident critical. The package continued to provide the framework, so an installation could appear to work normally.
According to timelines from StepSecurity, version 1.26.1 was published at 20:35 UTC and quarantined at 21:06 UTC. A second version, 1.26.2, appeared at 21:08 UTC and was quarantined at 21:16 UTC. PyPI’s current version is reported as 1.26.0, released on November 12. Both malicious releases have been removed from availability on the registry.
The advisory says the malicious code ran when Python started in an affected environment, without requiring a user to import the framework. Researchers reported collection of credentials and tokens available to developer machines and build environments. StepSecurity said it found no remote-code-loading second stage in the code it examined; the source material also reports that the malware did not persist beyond the installed package. Those findings do not establish whether any stolen credentials were used.
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
Credential Exposure in Build Pipelines
The incident matters because developer computers, continuous-integration systems and cloud build servers can hold credentials that grant access to source code, cloud services, package registries and AI platforms. A package that runs during Python startup can expose those secrets even if a developer never directly uses the affected framework. The advisory’s recommended response is to treat environments that installed either version as exposed, rebuild them from a clean base using 1.26.0, and rotate credentials they could access, including CI and GitHub Actions secrets.
The short interval between the two uploads also shows why removing one malicious release may not be enough if the publishing account or credential remains compromised. StepSecurity said the second release followed quarantine of the first by just over two minutes. JFrog likewise warned that quarantining a version does not itself revoke a compromised publishing credential or prevent another upload. That makes publisher access and registry response part of the incident, alongside cleanup by package users.
Wiz describes the package as receiving almost 200,000 downloads a month. That is a stated monthly download figure, not a count of affected installations or victims. The available source material does not establish how many downloads involved the two malicious releases, how many systems installed them, or whether credentials were subsequently abused.
Products Worth Considering

Amazon Kindle Scribe Colorsoft 64GB with Premium Pen

Bose QuietComfort Wireless Noise Cancelling Headphones, Sandstone

Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
As an affiliate, we earn on qualifying purchases.
How the Two Releases Appeared
Microsoft’s Agent Framework is an open-source Python toolkit for building AI agents and coordinating them in workflows. The package at issue, agent-framework, is an umbrella package that brings in framework components and integrations. The source material says Microsoft published the last clean version, 1.26.0, on November 12.
On November 17, the attacker published 1.26.1, described by the advisory as a wheel-only release. PyPI quarantined it later that evening, but StepSecurity reports that a second release was uploaded at 21:08 UTC, after the first had been quarantined at 21:06. PyPI quarantined the second version eight minutes later. The OpenSSF malware database listed the case as MAL-2026-20101 on November 17, and the GitHub Advisory Database published GHSA-764m-3ppc-5m73 on November 18.
Several security companies, including Socket, StepSecurity, Wiz, JFrog and Endor Labs, analyzed the releases. Their reports and the advisories provide the basis for the incident timeline and response guidance. The source material says the earlier package releases examined by Endor Labs included source code, while 1.26.1 was published as a wheel only; that difference is a reported observation, not by itself proof of how the attacker obtained publishing access.
“The malicious code ran “without importing agent_framework.””
— GitHub Advisory Database
Victim Count and Credential Use
The number of affected installations and organizations is not established in the supplied reporting. The monthly download figure cited by Wiz does not show how many people downloaded either malicious release, and no confirmed victim count is given. It is also unclear whether any stolen credentials were used after collection, which accounts or publishing credentials were compromised, and how the attacker obtained publishing access.
Microsoft’s cited statement says it made some repositories private while investigating malicious content, but the source material does not identify them or provide a fuller public incident account. Researchers’ findings about the inspected code, including the absence of a remote-code-loading second stage, describe what they examined; they do not prove that every potentially affected environment or consequence has been identified.
Rebuild, Rotate, and Verify
Organizations should first check package records and build logs for installations of agent-framework 1.26.1 or 1.26.2 during the November 17 exposure window. The advisory recommends treating any environment that installed either release as exposed, rebuilding it from a clean base on version 1.26.0, and rotating every credential the environment could access. That includes secrets available to CI systems and GitHub Actions, as well as relevant cloud, AI-service and package-registry credentials.
Security teams should also review relevant account and service logs for unusual credential activity and follow the technical advisories for detailed detection guidance. PyPI has quarantined the two releases, but that action alone does not resolve possible exposure on systems that already installed them. The next public developments to watch for are a fuller statement from Microsoft, any additional registry or vendor findings, and evidence clarifying the scope of installations or misuse.
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 in this incident.
Could the code run if I did not use the framework?
According to the advisory, the code ran when Python started in an affected environment and did not require importing agent_framework. Installation of either affected version is the key exposure condition described in the source material.
What should an organization do if it installed one of the releases?
Treat the environment as exposed, rebuild it from a clean base using version 1.26.0, and rotate credentials it could access, including CI and GitHub Actions secrets. Teams should use the advisories for technical response details.
How many people were affected?
No confirmed victim or installation count is provided in the source material. Wiz’s estimate of almost 200,000 downloads a month describes the package’s monthly downloads generally, not downloads of the malicious releases.
Did Microsoft explain how the attacker gained access?
Not in the statement cited by the source material. Microsoft told The Register it had made some repositories private while investigating malicious content; the method of access and fuller response were not specified.
Source: ThorstenMeyerAI.com
