🔍 Read the full analysis: The Microsoft AI Agent Framework PyPI Hijack: Who Needs To Respond? 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 Python package agent-framework contained credential-stealing code, according to security advisories. PyPI quarantined both releases, but anyone who installed either should treat the environment as exposed, rebuild it from a clean base and rotate credentials it could access. The number of affected installations and whether credentials were used remain unknown.
Two releases of Microsoft’s Agent Framework Python package on PyPI contained a credential stealer, according to the GitHub Advisory Database and security researchers. Versions 1.26.1 and 1.26.2 were published on November 17 and later quarantined; developers who installed either should treat the environment as exposed and rotate credentials it could reach.
The package, named agent-framework, is the package Microsoft’s installation instructions direct users to obtain from PyPI, the public registry for Python software. The GitHub Advisory Database says the two releases contained a credential stealer absent from the last clean release, version 1.26.0, and rates the incident critical. The source material says the package receives almost 200,000 downloads a month, citing Wiz; that figure is not a count of affected installations.
According to the advisory timeline compiled by researchers, version 1.26.1 was published at 20:35 UTC on November 17. PyPI quarantined it at about 21:06 UTC, according to StepSecurity. A second release, 1.26.2, appeared at 21:08 UTC and was quarantined at 21:16 UTC. PyPI’s current package version is reported as 1.26.0, released November 12. The releases are no longer available from the registry.
The advisory says the code ran when Python started in an affected environment, without requiring a user to import the framework. It sought credentials and tokens accessible to developer machines and build systems, including cloud, source-control and AI-service credentials. The reported response is to rebuild affected systems from a clean base using 1.26.0 and rotate every credential the environment could access, 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 Added Risk
The incident matters because the package could run in developer laptops, CI pipelines, build servers and container environments, where credentials may grant access to source code, cloud infrastructure, software publishing or AI services. A compromised token can create risks beyond the machine where it was stored, depending on its permissions and whether it remains valid.
The framework reportedly continued working normally, so a successful installation would not necessarily have produced an obvious error. The advisory says the code could run at Python startup, meaning later routine Python activity in an affected environment could trigger it. This makes checking whether a vulnerable version was installed more useful than relying on visible application symptoms.
Researchers also highlighted that quarantining the first release did not immediately prevent the second upload. StepSecurity said the attacker retained publishing access after the initial quarantine; JFrog similarly noted that removing one version does not revoke a compromised publishing credential. That sequence raises a separate response issue: registry action against a release and removal of an attacker’s ability to publish are distinct steps.
Popular right now

Ninja Foodi 6-in-1 DualZone FlexBasket Air Fryer, 7-Qt MegaZone, Black, DZ071

Vitamix VX1 Professional-Grade Blender, 64 oz, Black

Breville Barista Express Manual Espresso Machine with Built-In Grinder, Brushed Stainless Steel
As an affiliate, we earn on qualifying purchases.
How the Releases Reached PyPI
Microsoft’s Agent Framework is an open-source toolkit for building AI agents in Python and coordinating them in workflows. The agent-framework package is described as an umbrella package that pulls in the framework’s core code and integrations. Before the incident, the last release identified as clean in the advisory was 1.26.0, published November 12.
Security firms including Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases. The OpenSSF’s open-source malware database listed the case as MAL-2026-20101 on November 17, and the GitHub Advisory Database published GHSA-764m-3ppc-5m73 early on November 18. These records provide a public basis for the incident timeline and defensive guidance.
The first malicious release was uploaded as a wheel, a ready-to-install package file. StepSecurity reported that the second upload followed shortly after the first was quarantined and included renamed files and additional capabilities. The source material says researchers found no remote-code-loading second stage in the inspected code and describes the malware as not persisting beyond the installed package. Those findings do not establish whether credentials were taken or misused.
“The two releases contained a credential stealer that is not present in the last clean release, 1.26.0.”
— GitHub Advisory Database
Victim Counts and Credential Use
The available source material does not give a verified number of affected installations or organizations. Wiz’s estimate of almost 200,000 monthly downloads describes package traffic, not downloads of the malicious versions or confirmed victims. It is also not clear how many users installed either release during the short availability windows, including through automated updates.
No public information in the supplied material confirms whether stolen credentials were used, which accounts or services may have been accessed, or whether any downstream systems were affected. Microsoft’s brief statement does not describe the suspected publishing-access compromise, the repositories it made private or the scope of its investigation. The researchers’ analysis of the inspected code also cannot establish that no credentials were successfully exfiltrated.
Actions for Maintainers and Users
Organizations should check package records, lockfiles, build logs and deployment histories for versions 1.26.1 and 1.26.2, including installations made indirectly or by automation. Where either version was installed, the advisory-backed response is to rebuild from a clean base on 1.26.0 and rotate all credentials the affected environment could access. Teams should also review relevant service and CI logs for unusual use of those credentials and follow their incident-response procedures.
Microsoft’s investigation and any fuller company advisory could clarify how publishing access was obtained, what repositories were made private and whether the issue reached users beyond the known package releases. Further information from PyPI, Microsoft and the security firms may also establish the scale of exposure. Until then, the number of affected users and any confirmed misuse remain open questions.
Key Questions
Which versions of agent-framework were affected?
The GitHub Advisory Database identifies 1.26.1 and 1.26.2 as malicious. The last clean version identified in the advisory is 1.26.0.
What should a user do if either version was installed?
Treat the environment as exposed. Rebuild it from a clean base using 1.26.0 and rotate credentials the environment could access, including CI and GitHub Actions secrets.
Did the package have to be imported for the code to run?
No. The advisory says the malicious code ran when Python started in an affected environment, without requiring an import of agent-framework.
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 a general monthly package-download figure, not a count of downloads of the malicious releases.
Has Microsoft confirmed whether credentials were misused?
Not in the supplied material. Microsoft told The Register it had made some repositories private while investigating malicious content, but no confirmed credential misuse or investigation outcome is reported.
Source: ThorstenMeyerAI.com
