Microsoft’s AI Agent Framework Was Targeted On PyPI. What Happens Next?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Microsoft’s AI Agent Framework Was Targeted On PyPI. What Happens Next? on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

PyPI versions 1.26.1 and 1.26.2 of Microsoft’s agent-framework package contained credential-stealing code, according to security advisories. PyPI quarantined both releases, but the number of affected installations and any stolen credentials remain unknown; users who installed either version are advised to rebuild and rotate credentials.

Two releases of Microsoft’s Agent Framework Python package on PyPI contained credential-stealing code, according to the GitHub Advisory Database and security researchers. Versions 1.26.1 and 1.26.2 were uploaded on November 17 and quarantined within minutes, but developers who installed either version should treat the environment as exposed and rotate credentials it could access.

The GitHub Advisory Database says the malicious code was absent from the last clean release, 1.26.0, and rates the incident critical. PyPI records cited by the source show version 1.26.1 was published at 20:35 UTC and version 1.26.2 at 21:08 UTC on November 17. StepSecurity says PyPI quarantined the first at 21:06 UTC and the second at 21:16 UTC. Both versions are no longer available from PyPI; the current listed version is 1.26.0, released November 12.

The code was added through a Python startup hook, so it could run when Python started in an affected environment even if the user did not import Agent Framework, the advisory says. The package continued to function normally, meaning users might not have noticed anything wrong. Researchers say the code searched for credentials in environment variables and common developer-tool files, and sought cloud identity tokens. The second release added GitHub Actions identity-token access, according to the source material.

Security advisories tell users who installed either affected version to treat the relevant laptop, build server, CI pipeline or container as exposed. The recommended response is to rebuild from a clean base using version 1.26.0 and rotate every credential the environment could access, including CI and GitHub Actions secrets. The guidance applies even if the software was installed through an automated update.

At a glance
updateWhen: Incident occurred November 17, 2026; bo…
The developmentTwo compromised releases of Microsoft’s Python Agent Framework package were uploaded to PyPI on November 17, 2026, and later quarantined.

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

Risks for Developer and CI Credentials

The incident put credentials used to access cloud services, source-code repositories and AI platforms at potential risk. Such credentials can grant access beyond the machine where they were stored, including to build pipelines and hosted infrastructure. The source material does not establish that credentials were used after collection, but exposed tokens may remain usable until revoked or rotated.

The package’s normal operation made detection harder: a successful import or test would not necessarily reveal the malicious behavior. PyPI quarantining the releases stopped new installations through the registry, but it did not clean environments that had already installed them. That leaves developers and security teams with a response task even after the package versions disappeared.

Philips 5500 Series Fully Automatic Espresso Machine with LatteGo Milk System

Philips 5500 Series Fully Automatic Espresso Machine with LatteGo Milk System

Breville Barista Express Impress Espresso Machine with Grinder, Brushed Stainless Steel

Breville Barista Express Impress Espresso Machine with 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 Python toolkit for building AI agents and coordinating them in workflows. The PyPI package agent-framework serves as an umbrella package for framework components and integrations. The source says Microsoft’s own installation instructions pointed users to this package. Security firm Wiz described it as receiving almost 200,000 downloads a month; that figure is a reported monthly download estimate, not a count of affected installations.

Version 1.26.1 appeared as a wheel-only release, unlike earlier releases examined by Endor Labs, which included source code. After PyPI quarantined it, an attacker uploaded 1.26.2 just over two minutes later under renamed files, according to StepSecurity. Researchers from Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases. The OpenSSF malware database and GitHub Advisory Database subsequently published records of the case.

Researchers cited in the source found no remote-code-loading second stage in the inspected code and described the malware as not persistent beyond the installed package. That does not remove the credential risk: collected secrets may remain exposed after the package is removed, which is why advisories emphasize credential rotation as well as rebuilding.

“We have made some repositories private while we investigate malicious content.”

— Microsoft, in a statement to The Register

Victim Count and Credential Exposure

The source material gives no confirmed number of affected users or installations. Wiz’s estimate of almost 200,000 downloads a month does not show how many people installed versions 1.26.1 or 1.26.2 during the exposure window. It is also not clear whether attackers accessed or used any stolen credentials, or which organizations may have been affected.

Microsoft had said it made some repositories private while investigating malicious content, but the source reports no detailed Microsoft security advisory. The full circumstances behind the publishing access, including how it was compromised and whether other repositories or credentials were involved, remain unclear in the material provided.

Rebuild, Rotate, and Await Findings

Developers and organizations should identify whether their machines, images or pipelines installed either affected release, then follow the advisory guidance: rebuild from a clean base on version 1.26.0 and rotate credentials the environment could reach. Teams should include cloud, repository, package-registry and CI credentials in that review. Removing the package alone is not a substitute for revoking potentially exposed secrets.

Further details may come from Microsoft, PyPI or the security researchers as their investigations continue. Readers should check the GitHub Advisory Database and relevant organizational security channels for updated guidance. The available source does not give a timeline for a fuller Microsoft report or say whether PyPI and Microsoft have completed their investigations.

Key Questions

Which package versions were affected?

The reported affected PyPI releases are agent-framework 1.26.1 and 1.26.2. The advisory identifies 1.26.0 as the last clean release.

Could the malicious code run without importing Agent Framework?

Yes. The GitHub Advisory Database says a Python startup hook could run when Python started in an affected environment, even if the user did not import the framework.

What should someone do if they installed an affected version?

Treat the environment as exposed, rebuild it from a clean base using version 1.26.0, and rotate credentials it could access, including CI and GitHub Actions secrets.

How many users were affected?

No confirmed victim or installation count is provided in the source material. Wiz’s estimate of nearly 200,000 downloads per month is not a count of installs of the two malicious versions.

Has Microsoft published a detailed security advisory?

The source reports a short Microsoft statement to The Register about making some repositories private while investigating malicious content, but says it found no detailed Microsoft security advisory at the time of publication.

Source: ThorstenMeyerAI.com

You May Also Like

Which AI Automation Tools Should Small Businesses Consider?

Zapier offers easier setup and broad app connections; Make provides more control for complex workflows. Compare fit, cost and oversight needs.

Enhance Customer Engagement With Pre-Call Memory Cards In Sales

Testing of pre-call memory cards for relationship-driven sales professionals aims to improve client interactions by capturing human context beyond CRM data.

Review response quality coach for local service businesses

A new review response quality coach for local service businesses is being tested to improve reply professionalism, compliance, and efficiency.

Mobilisiert, nicht ausgegeben: Was von Europas €200-Milliarden-KI-Offensive übrig bleibt

Die EU kündigt eine KI-Investitionsoffensive mit 200 Mrd. € an, doch nur ein Bruchteil ist garantiert. Die tatsächlichen Mittel und Auswirkungen bleiben unklar.