🔍 Read the full analysis: Who Should Coordinate A Response To The AI Agent Framework PyPI Hijack? on ThorstenMeyerAI.com
Get hardware and tech essentials 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
PyPI quarantined Microsoft’s agent-framework versions 1.26.1 and 1.26.2 after security researchers identified credential-stealing code in both releases. The source reports that Microsoft is investigating and has made some repositories private; it does not identify a public lead coordinator or confirm how many environments were affected. Users of either version should treat exposed environments as compromised, rebuild from a clean base and rotate accessible credentials.
Two malicious versions of Microsoft’s Python package for Agent Framework were published on PyPI on November 17 and later quarantined, according to security advisories and researchers. The code was designed to collect credentials from affected environments, and while Microsoft says it is investigating, the available reporting does not identify a public, unified response coordinator or establish how many users were exposed.
The affected releases were agent-framework 1.26.1 and 1.26.2. The GitHub Advisory Database says both contained a credential stealer absent from the last clean version, 1.26.0, and rates the incident critical. The advisory says the malicious hook could run when Python started in an affected environment, even if a user did not import the Agent Framework package. That could make routine activity such as running a test or build script relevant to exposure.
According to the timeline reported by StepSecurity, version 1.26.1 appeared at 20:35 UTC on November 17 and was quarantined at 21:06. A second release, 1.26.2, followed at 21:08 and was quarantined at 21:16. PyPI’s current version is reported as 1.26.0, released November 12. The source says the second upload came shortly after the first was quarantined, highlighting that removal of a malicious version did not, by itself, stop another upload.
Researchers at Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases. The advisory describes collection of credentials and tokens from developer and cloud environments, including CI systems. StepSecurity reported finding no remote-code-loading second stage in the code it inspected. The source recommends treating any environment that installed either affected version as exposed, rebuilding from a clean base using 1.26.0, and rotating every credential the environment could access. Those are incident-response recommendations, not confirmation that every accessible credential was stolen.
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 Environments Face Added Risk
The incident matters because a package can remain functional while carrying malicious code. The source says the Agent Framework itself continued to work, so ordinary checks for a visibly broken installation might not alert users. If the reported startup behavior is accurate, an affected package could execute during later Python activity, not only when developers deliberately use the framework.
Build servers and CI pipelines can hold credentials with access to source code, cloud services, package registries or deployment systems. The advisory specifically calls for rotating credentials that were reachable from an affected environment, including CI and GitHub Actions secrets. The source does not establish that all those credentials were accessed or misused; it describes the capabilities researchers found and the prudent scope of response.
The sequence also makes coordination a practical security issue. PyPI can quarantine releases, researchers can analyze samples, and Microsoft can investigate its publishing systems and notify users. But the source does not say who is directing those efforts across organizations, whether a compromised publishing credential was revoked, or whether a central incident notice consolidates instructions. That gap can leave developers uncertain about which guidance to follow and whether they have completed remediation.
Products Worth Considering

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Apple MacBook Air (2019) 13-inch, 1.6GHz Intel Core i5, 8GB RAM, 128GB SSD, Gold
As an affiliate, we earn on qualifying purchases.
How the Two Releases Reached PyPI
Microsoft’s Agent Framework is an open-source toolkit for building AI agents and workflows in Python. The PyPI package agent-framework is described in the source as an umbrella package that installs the framework core and integrations; it is the package referenced by Microsoft’s installation instructions. Wiz said it receives “almost 200k downloads a month,” but the source gives no measurement window or baseline for that figure, and it should not be read as a count of affected installations.
The documented timeline begins with clean version 1.26.0 on November 12. The advisory says 1.26.1 was published as a wheel only; Endor Labs said the releases it examined from 1.21.0 through 1.26.0 had also included source code. After PyPI quarantined 1.26.1, the attacker uploaded 1.26.2. StepSecurity said quarantining the first version did not immediately revoke the attacker’s publishing access, and JFrog likewise warned that quarantining a release does not itself revoke a compromised publishing credential.
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, according to the source. These advisories and independent research give defenders technical and remediation information. The source says Microsoft had not published a security advisory as of November 19, while The Register reported a brief statement from the company.
“We have made some repositories private while we investigate malicious content.”
— Microsoft, in a statement to The Register, as quoted by ThorstenMeyerAI.com
Open Questions in the Response
The number of affected environments is unknown. The source says no victim count had been published as of November 19. A download estimate for the package cannot establish how many installations occurred during the exposure window, how many ran Python afterward, or whether any credentials were successfully used.
It is also unclear from the available reporting whether Microsoft identified and revoked the publishing credential or other access used for the uploads, what repositories were made private, and whether the company will issue a detailed advisory. Researchers’ findings about the inspected releases do not prove that no other activity occurred outside the code they analyzed. The source contains no confirmed account of downstream intrusions, data theft beyond credential collection, or subsequent misuse of stolen tokens.
The question in the headline—who should coordinate—therefore remains partly unanswered by the cited material. Microsoft is the package maintainer and says it is investigating; PyPI quarantined the releases; security researchers and advisory databases documented the incident. The source does not report a designated incident commander, a joint response channel, or a single authoritative public update bringing those roles together.
What Maintainers and Users Need Next
For users, the immediate steps in the advisory are to identify systems that installed 1.26.1 or 1.26.2, rebuild those environments from a clean base using the clean 1.26.0 release, and rotate credentials the environments could access. Teams should include laptops, CI workers, containers and automated dependency-update processes in that review. The source does not provide indicators of compromise in its article; defenders should consult the referenced advisories for technical details.
For the response to become easier to follow, Microsoft would need to publish an account of the package and publishing-system investigation, clarify the status of publisher access, and state what users should check beyond reinstalling and rotating secrets. PyPI and security researchers can provide their own findings, but a clear maintainer update would help users distinguish confirmed facts from precautionary actions. The available source does not say when such an update will arrive.
Until more information is published, teams should treat the incident as ongoing and incompletely quantified, rather than infer that quarantine alone resolved exposure. Further advisories, a Microsoft incident notice, or evidence of affected downstream systems could change the assessment; none is confirmed in the source material as of November 19.
Key Questions
Which package versions were affected?
The affected PyPI releases were agent-framework 1.26.1 and 1.26.2. The source identifies 1.26.0, released November 12, as the last clean version.
Who should coordinate the response?
The source does not name a designated coordinator. Microsoft is investigating as the package maintainer, PyPI quarantined the releases, and security firms and advisory databases documented the issue. A public, unified response lead is not confirmed.
What should a team do if it installed an affected release?
The cited advisory recommends treating the environment as exposed, rebuilding it from a clean base on version 1.26.0, and rotating every credential it could access, including relevant CI and GitHub Actions secrets. Check the advisories for technical response details.
How many users or systems were affected?
No victim count is reported in the source material. Package download estimates do not establish how many installations occurred during the affected window or whether credentials were stolen or used.
Has Microsoft published a detailed security advisory?
As of November 19, the source says it had found no Microsoft security advisory. It reports that Microsoft told The Register it had made some repositories private while investigating malicious content.
Source: ThorstenMeyerAI.com
