AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Microsoft’s AI Agent Framework Was Hijacked On PyPI: Next Steps For Users on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Two versions of Microsoft’s Python Agent Framework package, 1.26.1 and 1.26.2, contained credential-stealing code and were briefly available on PyPI on Nov. 17, 2026. Security advisories say users who installed either version should treat the environment as exposed, rebuild it using a clean release, and rotate credentials it could access.

Two releases of Microsoft’s Agent Framework Python package carried credential-stealing code while briefly available on PyPI on Nov. 17, 2026, according to security advisories. PyPI quarantined versions 1.26.1 and 1.26.2, but users who installed either should treat the affected environment as exposed and take steps to rebuild it and replace accessible credentials.

The affected package is agent-framework, the official PyPI package that Microsoft’s installation instructions point to. The GitHub Advisory Database says releases 1.26.1 and 1.26.2 contained a credential stealer absent from the last clean release, version 1.26.0. The advisory rates the incident critical. The framework continued to work, making the malicious additions less likely to be noticed through ordinary use.

The releases appeared at 20:35 UTC and 21:08 UTC, respectively. StepSecurity says PyPI quarantined version 1.26.1 at 21:06 UTC and version 1.26.2 at 21:16 UTC. The second upload followed the first quarantine by just over two minutes. PyPI’s current version is reported as 1.26.0, released Nov. 12; both malicious releases have been removed from availability.

The advisory says a startup hook caused the code to run when Python started in an affected environment, without requiring the user to import the framework. Researchers reported collection of credentials and tokens from developer tools, cloud services and build environments. StepSecurity said its inspection found no remote-code-loading second stage, and The Register reported researchers’ assessment that the malware did not persist beyond the installed package. Those findings do not establish whether stolen credentials were used.

At a glance
updateWhen: Incident occurred Nov. 17, 2026; adviso…
The developmentTwo malicious versions of Microsoft’s Agent Framework package were published to PyPI before the registry quarantined them on Nov. 17.

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

  1. Clean 1.26.0 publishedStill the current version on PyPI
  2. Malicious 1.26.1 appearsFramework works normally, a credential stealer runs whenever Python starts
  3. PyPI quarantines 1.26.1After about 31 minutes
  4. Attacker uploads 1.26.2Renamed files, also targets GitHub Actions tokens
  5. PyPI quarantines 1.26.2Eight minutes later
  6. Malware databases list itMAL-2026-20101 (OpenSSF), GHSA-764m-3ppc-5m73 (GitHub)
Quarantine is not revocation. The second release came 2 minutes 23 seconds after the first was pulled: the attacker still had publishing access. Copies already in caches, mirrors and images stay there.

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 Credentials Were at Risk

The incident matters because the package could be installed on more than a developer’s personal laptop. CI pipelines, build servers and container images may hold credentials that can access source repositories, cloud accounts, deployment systems or AI services. A compromised environment could expose those secrets even if the developer’s application kept running normally.

Security advisories recommend treating any environment that installed either affected version as exposed. That means removing and rebuilding the environment from a clean base, then rotating every credential it could reach, including CI and GitHub Actions secrets. Simply uninstalling or downgrading the package does not address a credential that may already have been copied elsewhere.

The two uploads also show why quarantine of a specific release may not end an incident. StepSecurity and JFrog said quarantining the first version did not immediately revoke the attacker’s publishing access; another release followed. The timeline highlights the need to verify package publisher access as well as remove malicious files.

Products Worth Considering

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

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 and coordinating workflows in Python. The PyPI package agent-framework is an umbrella package that brings in the core framework and integrations. Wiz says the package receives almost 200,000 downloads a month, though that figure does not establish how many installations occurred during the incident window.

According to the GitHub Advisory Database and researchers’ timelines, Microsoft’s last clean release, 1.26.0, was published Nov. 12. Version 1.26.1 appeared five days later as a wheel-only release. PyPI quarantined it, but version 1.26.2 was uploaded shortly afterward with renamed files and additional capabilities, according to StepSecurity. PyPI quarantined that second release eight minutes after its publication.

The OpenSSF’s malware database listed the case as MAL-2026-20101 on Nov. 17, and the GitHub Advisory Database published GHSA-764m-3ppc-5m73 early on Nov. 18. Security firms including Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases. Wiz said it reported its findings to Microsoft on Nov. 17.

“The credential stealer is not present in the last clean release, 1.26.0.”

— GitHub Advisory Database

Victim Count and Credential Use

The number of affected installations and victims has not been established in the supplied reporting. A package’s reported monthly download volume is not a count of downloads during the brief incident window, nor does it show how many users installed the malicious versions.

It is also not clear whether stolen credentials were used to access accounts or systems, whether any data or services were affected, or how the attacker obtained the publishing access. Microsoft’s reported statement says repositories were made private during an investigation, but the material provides no full incident report or detailed account of the response. Security researchers’ code findings describe what the releases could do; they do not confirm successful theft from a particular victim.

Actions for Package Users

Organizations and individual developers should check dependency records, lockfiles, build logs and package caches for agent-framework 1.26.1 or 1.26.2. If either was installed, the advisory’s recommended response is to treat that environment as exposed, rebuild it from a clean base using the clean 1.26.0 release, and rotate credentials the environment could access. Teams should also review relevant account and service logs for unexpected activity.

Security teams should follow the GitHub and OpenSSF advisories for technical indicators and response guidance; this report does not reproduce indicators or attack instructions. Microsoft’s investigation and any further findings about the publishing account, affected users or credential misuse were not available in the supplied material. Users should watch for additional advisories and verify the package version before installing or updating.

Key Questions

Which package versions were affected?

The affected PyPI releases were agent-framework 1.26.1 and 1.26.2. The GitHub Advisory Database identifies 1.26.0 as the last clean release in the reported sequence.

Could the malicious code run without importing the framework?

Yes. The advisory says a startup hook could run when Python started in an affected environment, without importing agent_framework.

What should someone do after installing an affected release?

Treat the environment as exposed, rebuild it from a clean base using a clean release, and rotate all credentials it could access, including CI and GitHub Actions secrets. Check logs for suspicious activity as part of the response.

How many users were affected?

The supplied reporting does not give a confirmed victim or installation count. Wiz’s estimate of almost 200,000 monthly downloads is a broader package figure, not a count of downloads during the incident.

Has Microsoft confirmed that credentials were misused?

No such confirmation appears in the supplied material. Microsoft told The Register it had made some repositories private while investigating malicious content; whether credentials were used or accounts were accessed remains unclear.

Source: ThorstenMeyerAI.com

You May Also Like

Cloudflare Brings OpenAI Into Its Vulnerability Discovery And Defense Efforts

Cloudflare says it is working with OpenAI on AI-assisted vulnerability detection, but has not disclosed deployment scope or performance results.

Could Robots Benefit From Their Own AI? OpenAI And Runway Raise The Issue

OpenAI says general models can handle robot control as Runway releases an open specialist model. Early results show trade-offs in success and speed.

How AI Is Framed In Government Workforce Debates Around The World

A Dutch government savings plan and workforce cuts elsewhere renew debate over AI, but documents show budgets often drive reductions more directly.

Claude Fable 5.5: Anthropic Gets Ready For A Next-Week Release

TestingCatalog reports a possible Claude Fable 5.5 release next week, but Anthropic has not confirmed the model or timing in available information.