🔍 Read the full analysis: Microsoft’s AI Agent Framework PyPI Hijack: Who Owns The Response? on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Versions 1.26.1 and 1.26.2 of Microsoft’s PyPI package agent-framework contained credential-stealing code, according to the GitHub Advisory Database and security researchers. PyPI quarantined both releases on November 17, 2026; anyone whose environment installed either version should treat it as exposed, rebuild from a clean base and rotate accessible credentials.
Two releases of Microsoft’s Agent Framework Python package on PyPI contained credential-stealing code, according to the GitHub Advisory Database and multiple security firms. The versions, 1.26.1 and 1.26.2, were published and quarantined on November 17, 2026; developers and organizations that installed either should treat affected environments as exposed and rotate credentials the software could access.
The affected package was agent-framework, the official PyPI package linked from Microsoft’s installation instructions, according to the source report. The GitHub Advisory Database says the two releases included a credential stealer absent from the previous clean release, version 1.26.0, and rates the incident critical. The package’s framework functionality continued to work, so an installation did not necessarily produce an obvious error.
According to the advisory, a startup hook caused the malicious code to run when Python started in an affected environment, even if a user did not import the framework. Researchers said it sought credentials and tokens accessible to developer machines and build systems, including cloud, source-control and AI-service credentials. The advisory also describes requests for fresh tokens from cloud and pipeline identity services. These findings describe the code’s capabilities; the source material does not establish that every targeted credential was successfully obtained.
The timeline reported by StepSecurity places publication of version 1.26.1 at 20:35 UTC on November 17 and its quarantine at 21:06. Version 1.26.2 appeared at 21:08 and was quarantined at 21:16. PyPI’s current listed version was 1.26.0, released November 12, according to the source. The advisory recommends rebuilding affected environments from a clean base and rotating credentials they 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 Systems Face Exposure
The incident matters because the package was distributed through the official public registry and the malicious code reportedly ran at Python startup. A user did not need to deliberately call a suspicious feature: later routine activity in an affected environment could trigger the hook. That creates potential exposure across laptops, CI runners, build servers and container images that installed one of the releases.
Build environments can hold credentials that grant access to cloud resources, code repositories and deployment systems. Security researchers cited in the source material said the code sought both stored secrets and short-lived identity tokens. If credentials were accessible and obtained, consequences could extend beyond a developer’s machine. The available reporting does not confirm how many environments were affected or whether stolen credentials were used.
The short gap between the two uploads also puts attention on publishing access and incident response. StepSecurity and JFrog said quarantining the first version did not, by itself, revoke the compromised publishing credential or prevent another release. That observation does not establish how Microsoft’s credentials were compromised, but it shows why removing a malicious package and stopping further publication are separate response tasks.
Popular right now

Samsung 65-Inch 4K OLED Smart TV with Dolby Atmos, 120Hz Refresh Rate, and Amber Protection Bundle

Dyson V8 Cyclone Cordless Vacuum

Samsung 65-Inch S90H OLED 4K Smart TV Bundle with HW-B630F Soundbar, Subwoofer, and Screen Cleaner (2026)
As an affiliate, we earn on qualifying purchases.
Two Releases, One Publication Window
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 is described in the source material as an umbrella package that installs framework code and integrations. This makes the incident relevant to people who may have installed the package as a dependency rather than directly selecting an individual component.
The clean version identified in the reporting was 1.26.0, published November 12. Endor Labs said releases it examined from 1.21.0 through 1.26.0 also included source code, while the first malicious release was published as a wheel only. Following the two uploads and quarantines, the OpenSSF malware database and GitHub Advisory Database recorded the case. Security firms named in the source material include Socket, StepSecurity, Wiz, JFrog and Endor Labs.
Wiz said it promptly reported its findings to Microsoft on November 17. The article’s source reports no Microsoft security advisory at publication time. It quotes Microsoft telling The Register: “We have made some repositories private while we investigate malicious content.” The statement confirms an investigation-related repository change, but does not explain which repositories were made private or how the package’s publishing access was secured.
““We have made some repositories private while we investigate malicious content.””
— Microsoft, in a statement to The Register
Victim Count and Credential Use
The source material gives no confirmed victim count, download count for the two malicious versions, or estimate of how many installations occurred before quarantine. Wiz is cited as saying the package receives almost 200,000 downloads a month, but that figure is not a count of downloads of the affected releases and does not show how many users installed them.
It is also not clear whether attackers successfully collected credentials, accessed systems with them, or used any tokens after collection. Researchers reportedly found no remote-code-loading second stage in the code they inspected, and the source says the malware was not persistent beyond the installed package. Those findings do not rule out exposure during the period an affected package remained installed. The source material does not describe how the attacker obtained publishing access or confirm that Microsoft has completed its investigation.
Rebuild, Rotate, and Await Advisories
Organizations should use the security advisory’s guidance to identify environments that installed 1.26.1 or 1.26.2, rebuild them from a clean base using 1.26.0 or another verified safe release, and rotate every credential those environments could access. The recommendation includes CI and GitHub Actions secrets. Teams should assess accessible credentials and systems as potentially exposed rather than assuming that a working application was clean.
Further clarity depends on updates from Microsoft, PyPI and the researchers. Readers should look for a fuller account of the publishing-access compromise, the scope of affected installations, and any evidence of credential use. The source material provides no date for a Microsoft advisory or a completed investigation, so those developments remain pending.
Key Questions
Which package versions were affected?
The reported affected versions are agent-framework 1.26.1 and 1.26.2. The GitHub Advisory Database identifies 1.26.0 as the last clean release in the account provided.
Does installing the package prove that credentials were stolen?
No. The advisory and researchers describe credential-stealing capabilities, but the source material does not confirm that every installation exposed credentials or that stolen credentials were used. Environments that installed either affected version should still be treated as exposed.
What should an organization do if it installed an affected version?
Rebuild the environment from a clean base and rotate credentials it could access, including relevant cloud, repository and CI secrets. Follow the security advisory for detailed response steps.
How many users or systems were affected?
A confirmed victim or installation count is not provided. The source cites an estimate of almost 200,000 package downloads per month, but that is not a count of downloads or installations of versions 1.26.1 and 1.26.2.
What has Microsoft said publicly?
The source reports that Microsoft told The Register: “We have made some repositories private while we investigate malicious content.” It reports no Microsoft security advisory at the time of publication.
Source: ThorstenMeyerAI.com
