🔍 Read the full analysis: Microsoft’s AI Agent Framework PyPI Hijack: What Should Users Do? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Two malicious releases of Microsoft’s Python package agent-framework were briefly available on PyPI on November 17, 2026, and contained code that sought developer and cloud credentials. PyPI quarantined both releases; anyone whose environment installed either version should treat it as exposed, rebuild from a clean base and rotate credentials it could access.
Two releases of Microsoft’s agent-framework Python package, versions 1.26.1 and 1.26.2, were published to PyPI on November 17, 2026, with code that sought credentials from affected environments, according to security advisories and researchers. PyPI quarantined both releases, but developers and organizations whose machines installed either version should treat those environments as exposed and investigate credentials they could reach.
The GitHub Advisory Database says the two releases contained a credential stealer absent from the last clean release, 1.26.0, and rates the incident critical. The package is the official PyPI package referenced by Microsoft’s installation instructions, rather than an unrelated package using a similar name. Security firm Wiz says it receives almost 200,000 downloads a month; that figure is a reported monthly download rate, not a count of confirmed affected installations.
According to the advisory and researcher timelines, version 1.26.1 was published at 20:35 UTC on November 17. StepSecurity says PyPI quarantined it at 21:06 UTC. Version 1.26.2 appeared at 21:08 UTC and was quarantined at 21:16 UTC. The reported interval between the first release and its quarantine was about 31 minutes. PyPI’s current listed version is 1.26.0, released November 12, according to the source material.
The advisory says the malicious code ran when Python started in an affected environment, even if a user did not import the framework itself. It sought credentials and tokens available to developer machines and build systems, including cloud, CI and AI-service credentials. The advisory specifically tells affected users to consider secrets such as those used by CI and GitHub Actions exposed. The framework could continue working normally, so an obvious failure was not a reliable sign that an environment was clean.
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 Python packages are often installed automatically into laptops, continuous-integration jobs and container builds. A release that executes at interpreter startup can put credentials at risk during routine tasks, even when developers do not intentionally use the affected framework. That makes the potential exposure broader than the application that directly depends on the package.
Credentials available to a build environment can grant access to source code, cloud resources, package registries or deployment systems. The advisory describes collection of credentials from files and environment variables, as well as requests for certain cloud and pipeline identity tokens. OpenSourceMalware, as quoted by The Register, said the mix of credential collection and active token requests was especially concerning in cloud build environments because short-lived tokens can still be useful if sent out promptly. This is the group’s assessment, not a published count of stolen or misused credentials.
Researchers also highlighted a release-control issue: quarantining the first malicious version did not immediately prevent the second upload. StepSecurity and JFrog said the initial quarantine did not itself revoke the publishing access used to push another release. That sequence makes package-registry response and publisher credential security relevant beyond this single package.
Products Worth Considering

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

FLEXISPOT 3-in-1 Convertible Chair and Table for Four

Vari Ergo Electric Standing Desk, 54×26, Light Wood/White
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 acts as an umbrella package that brings in the framework’s core code and integrations. Its legitimate version 1.26.0 was published on November 12, according to the package timeline in the source material.
Researchers from Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the suspect releases. Endor Labs noted that releases it examined from 1.21.0 through 1.26.0 had included source code, while 1.26.1 was uploaded as a wheel only. StepSecurity reported that the second upload arrived just over two minutes after the first was quarantined, with renamed files and additional capabilities. Those findings describe differences researchers observed; they do not identify who carried out the attack.
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, according to the supplied timeline. Wiz said it promptly reported its findings to Microsoft on November 17. The source material says Microsoft had made a short statement to The Register: “We have made some repositories private while we investigate malicious content.”
“The two releases contained “a credential stealer that is not present in the last clean release, 1.26.0.””
— GitHub Advisory Database
Victim Numbers Remain Unknown
The source material does not provide a verified count of downloads during the exposure window, affected installations, compromised accounts or confirmed credential misuse. Wiz’s reported monthly download figure is not an incident-specific victim estimate. It is also unclear which organizations’ machines installed the releases or whether any stolen credentials were used after collection.
The source material says StepSecurity found no remote-code-loading second stage in the code it inspected, and reports that the malware was not persistent beyond the installed package. Those are findings about the inspected samples, not proof that every affected system is safe or that no separate activity occurred. Microsoft’s full investigation findings, the publishing-access compromise and its scope are not established in the material provided.
Rebuild and Rotate Credentials
Organizations should first check dependency lockfiles, package caches, build logs and container histories for agent-framework 1.26.1 or 1.26.2. Treat any environment that installed either release as exposed, including development machines, CI runners and images built during the relevant period. The advisory recommends rebuilding from a clean base using the clean 1.26.0 release rather than assuming that removing or downgrading the package alone clears an affected environment.
Security teams should identify credentials accessible to those environments and rotate them, including cloud keys, CI secrets, GitHub Actions credentials and AI-service tokens where applicable. Review relevant access and audit logs for unusual use, and update downstream images or caches so the suspect releases are not reintroduced. The advisory provides technical details for defenders; this article does not reproduce indicators of compromise or attack instructions.
Microsoft’s repositories and publishing controls remain part of the developing response described in the source material. Users should watch for a Microsoft security advisory and updates from PyPI, the GitHub Advisory Database and the researchers. Until investigators publish more findings, the number of affected users and any confirmed follow-on use of credentials remain unknown.
Key Questions
Which versions of agent-framework were affected?
The suspect PyPI releases were 1.26.1 and 1.26.2. The GitHub Advisory Database identifies 1.26.0 as the last clean release in this incident.
What should I do if I installed one of the affected versions?
Treat the environment as exposed. Rebuild it from a clean base using the clean release, then rotate credentials that the environment could access. Check CI systems, container images and developer machines, not only the application that directly uses the framework.
Does a working application mean my system was unaffected?
No. The advisory says the package could continue working normally while its malicious code ran when Python started. Normal application behavior does not establish that an environment was clean.
How many users were affected or had credentials stolen?
The supplied material gives no verified victim count and no confirmed count of stolen or misused credentials. Wiz’s estimate of almost 200,000 downloads a month is a general monthly figure, not a measure of installations during the incident window.
Source: ThorstenMeyerAI.com
