🔍 Read the full analysis: Who Should Address The Credential Theft Risk In Microsoft’s AI Agent Framework? on ThorstenMeyerAI.com
Get the latest gadgets 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 Python package agent-framework, 1.26.1 and 1.26.2, briefly carried credential-stealing code on PyPI on November 17, 2026. PyPI quarantined both releases, but anyone who installed either should treat accessible credentials as exposed and rebuild the environment from a clean version. The source material does not establish how many systems were affected or who compromised the package.
Two releases of Microsoft’s Agent Framework Python package briefly carried credential-stealing code after appearing on PyPI on November 17, according to security advisories and research firms. The releases, 1.26.1 and 1.26.2, were quarantined by the registry, but developers and organizations that installed either version are being advised to treat accessible credentials as exposed and rebuild affected environments.
The GitHub Advisory Database says the malicious code was absent from the last clean release, version 1.26.0, and rates the incident critical. PyPI records and researcher timelines put the first upload at 20:35 UTC and the second at 21:08 UTC on November 17. StepSecurity says PyPI quarantined the first release at 21:06 UTC and the second at 21:16 UTC. The source material says the current PyPI version is 1.26.0, published November 12.
The package was distributed under Microsoft’s name and is the package its installation instructions point to, according to the source report. Wiz says it receives almost 200,000 downloads a month; that figure is a reported monthly download count, not an estimate of affected installations. Several security firms, including Socket, StepSecurity, Wiz, JFrog and Endor Labs, examined the releases. Wiz said it reported its findings to Microsoft on November 17.
The advisory says the malicious code ran when Python started in an affected environment, even if the user did not import the Agent Framework. It sought credentials and tokens available to developer machines and build systems, including cloud, code-hosting and AI-service credentials. The advisory recommends treating environments that installed either version as exposed, rebuilding from a clean base on 1.26.0, and rotating every credential those environments 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
Credential Exposure Across Build Systems
The incident matters because software packages often run inside environments that hold credentials with access to source code, cloud infrastructure and deployment pipelines. A developer may install a dependency on a laptop, while an automated build can install it with access to CI secrets or cloud identity services. If a malicious package can read or request those credentials, the potential impact can reach beyond the machine where it was installed.
The advisory says the code could collect environment variables and credential files and, in some cases, request fresh tokens from cloud or pipeline identity services. OpenSourceMalware, as quoted by The Register, said the combination of passive theft and active token requests is particularly dangerous in cloud build environments because short-lived tokens can be useful to an attacker if exfiltrated quickly. That is a researcher’s assessment of the risk, not confirmation that stolen credentials were used.
The episode also highlights a distinction between removing a malicious release from a registry and containing a compromised publishing account. StepSecurity said quarantining the first version did not immediately revoke publishing access; JFrog similarly noted that quarantining a package version does not by itself prevent another upload. The quick appearance of version 1.26.2 after the first quarantine makes control of release access a central issue in the response.
Popular right now

TCL F8D Series 75-Inch 4K QLED Smart Fire TV, 144Hz, HDR (75F8D, 2026)

Vitamix 5200 Professional-Grade Blender with 64 oz Container, Black

Breville Barista Express Manual Espresso Machine with Built-In Grinder, Brushed Stainless Steel
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 named agent-framework acts as an umbrella package, bringing in the framework’s core code and integrations. That role means users could encounter the affected package through the documented installation route rather than through an unrelated lookalike.
The clean version 1.26.0 was published on November 12. Five days later, version 1.26.1 appeared as a wheel-only release, according to the GitHub advisory. Endor Labs said releases it examined from 1.21.0 through 1.26.0 had also included source code. After PyPI quarantined 1.26.1, a second release was uploaded about two minutes later, according to StepSecurity; the new version used renamed files and added capabilities, researchers reported.
OpenSSF’s open-source malware database listed the case as MAL-2026-20101, and the GitHub Advisory Database published GHSA-764m-3ppc-5m73. The source material says 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.
““a credential stealer that is not present in the last clean release, 1.26.0””
— The GitHub Advisory Database
Victim Count and Access Remain Unknown
The source material does not give a verified count of downloads or affected installations during the exposure window. Wiz’s figure of almost 200,000 downloads a month describes the package’s reported monthly volume, but it does not show how many people installed versions 1.26.1 or 1.26.2. It is also not clear how many organizations’ credentials were reachable from affected environments.
The material does not identify who carried out the compromise or explain how publishing access was obtained. It also does not establish whether stolen credentials were used after collection, whether any systems were accessed, or whether data beyond credentials was taken. StepSecurity reported no remote-code-loading second stage in the code it inspected, and The Register summarized researchers’ findings that the malware did not persist beyond the installed package; those observations do not answer whether credentials copied before removal were misused.
Microsoft’s reported statement says it is investigating malicious content, but the source report found no company security advisory at publication. The full scope of the investigation, the status of publishing credentials and any confirmed downstream effects remain unclear.
Rebuild, Rotate, and Await Findings
Organizations should first check dependency records, build logs and package lockfiles for installations of agent-framework 1.26.1 or 1.26.2. The advisory’s stated response is to treat any environment that installed either version as exposed, rebuild it from a clean base using 1.26.0, and rotate credentials the environment could reach. Replacing the package alone does not invalidate credentials that may already have been copied.
Teams should also review relevant cloud, code-hosting and CI account activity for signs of unauthorized use, while following the detailed guidance in the security advisories. The source article deliberately omits indicators of compromise and technical attack details, directing defenders to those advisories instead. Administrators should distinguish confirmed installation of an affected version from evidence that a credential was actually accessed or abused.
The next major developments are expected to come from Microsoft’s investigation and any further updates from PyPI, the GitHub Advisory Database, OpenSSF or the researchers. Those updates could clarify how publishing access was compromised, how many users installed the releases and whether credentials were misused. Until then, the confirmed exposure window and the package versions provide the clearest basis for response.
Key Questions
Which Microsoft Agent Framework versions were affected?
The affected PyPI releases were agent-framework 1.26.1 and 1.26.2. The source material identifies 1.26.0, published November 12, as the last clean release.
Did using the framework require importing it for the code to run?
No. The GitHub advisory says the malicious code ran when Python started in an affected environment, without importing agent_framework. A Python process launched later in that environment could trigger it.
What should a team do if it installed an affected release?
The advisory recommends treating the environment as exposed, rebuilding from a clean base on version 1.26.0, and rotating every credential it could access, including CI and GitHub Actions secrets.
How many people or organizations were affected?
No verified victim or affected-installation count is provided in the source material. Wiz’s estimate of almost 200,000 downloads per month is not a count of downloads of the malicious versions.
Has Microsoft explained how the package was compromised?
Not in the statement cited by the source material. Microsoft told The Register it had made some repositories private while investigating malicious content; the cause and status of the investigation were not reported there.
Source: ThorstenMeyerAI.com
