Who Is Expected To Respond To The Microsoft AI Agent Framework PyPI Attack?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Who Is Expected To Respond To The Microsoft AI Agent Framework PyPI Attack? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys 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 releases of Microsoft’s official Python package agent-framework, versions 1.26.1 and 1.26.2, contained credential-stealing code and were published on November 17, 2026. PyPI quarantined both, but anyone who installed either version should treat the environment and credentials it could access as exposed; Microsoft has said it is investigating, while further response details remain limited.

Two releases of Microsoft’s official Python package for its Agent Framework carried credential-stealing code before PyPI quarantined them on November 17, according to public security advisories. Developers and organizations that installed agent-framework 1.26.1 or 1.26.2 are being urged to rebuild affected environments and rotate credentials those systems could reach; Microsoft has said it is investigating malicious content in some repositories.

The GitHub Advisory Database identifies versions 1.26.1 and 1.26.2 as malicious and rates the incident critical. The releases were uploaded to PyPI, the public registry used to distribute Python packages. The advisory says the code added a startup hook that could run when Python started, even if a user did not import the Agent Framework. The framework continued to work, which could make the compromise less apparent to users.

The documented timeline shows version 1.26.1 appearing at 20:35 UTC on November 17. StepSecurity says PyPI quarantined it at 21:06 UTC; version 1.26.2 was uploaded at 21:08 UTC and quarantined at 21:16. The current PyPI release identified in the source material is 1.26.0, published November 12. Security researchers at Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases, and public records appeared in the OpenSSF malware database and GitHub Advisory Database.

The GitHub advisory says the code sought credentials including cloud keys, developer-tool credentials and tokens, and could access secrets available to the compromised environment. The described targets include credentials associated with Azure, AWS, GitHub, Hugging Face, OpenAI and package registries. Its remediation guidance is to treat systems that installed either affected version as exposed, rebuild from a clean base using 1.26.0, and rotate credentials the environment could access, including CI and GitHub Actions secrets.

At a glance
reportWhen: Incident occurred November 17, 2026; bo…
The developmentSecurity researchers and public advisories documented two malicious releases of Microsoft’s Agent Framework package on PyPI, prompting calls for affected users to rebuild environments and rotate credentials.

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 Environments Face Risk

The incident matters because a compromised developer laptop or automated build system can hold credentials that grant access well beyond the package itself. A stolen cloud key, CI token or AI service credential could potentially expose connected services or repositories. The actual reach depends on which credentials were present and what permissions they carried; the available source material does not establish that every targeted account or service was accessed.

The attack also illustrates a registry response problem: quarantining one release did not prevent a second upload. StepSecurity says the first quarantine did not immediately revoke the publisher’s access, and JFrog similarly noted that removing a version does not by itself invalidate a compromised publishing credential. For users, removing the package is not enough if secrets may already have been copied. Rebuilding and credential rotation address separate parts of the risk.

Wiz describes the package as receiving almost 200,000 downloads a month. That is a reported download rate, not a count of affected installations or victims. The number of systems that installed the malicious versions, and whether any stolen credentials were used, has not been established in the supplied reporting.

DREAME D20 Air Plus Robot Vacuum and Mop with 20,000 Pa Suction

DREAME D20 Air Plus Robot Vacuum and Mop with 20,000 Pa Suction

Philips 5500 Series Fully Automatic Espresso Machine with LatteGo Milk System

Philips 5500 Series Fully Automatic Espresso Machine with LatteGo Milk System

As an affiliate, we earn on qualifying purchases.

From Clean Release to Quarantine

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 framework code and integrations. Microsoft’s installation instructions reportedly pointed users to the official package, distinguishing this incident from a similarly named copycat.

Version 1.26.0 was published November 12. The next two releases arrived on November 17. The GitHub advisory notes that 1.26.1 was uploaded 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 the first release was quarantined, the second appeared roughly two minutes later, with changed file names and additional capabilities, according to StepSecurity.

Public documentation followed the quarantines. The OpenSSF malware database listed the case as MAL-2026-20101 on November 17, and GitHub published advisory GHSA-764m-3ppc-5m73 early November 18. The advisories and researchers’ findings are the basis for the current account; they do not by themselves identify how the publishing access was compromised.

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

— Microsoft, in a statement to The Register

Victims and Credential Use Unknown

The available source material does not give a confirmed count of downloads, affected installations or victims for versions 1.26.1 and 1.26.2. Wiz’s estimate of almost 200,000 monthly downloads concerns the package generally, not the number of downloads during the exposure window.

It is also unclear how the attacker obtained publishing access, whether credentials were successfully exfiltrated from specific systems, and whether any stolen tokens were used. StepSecurity reported finding no remote-code-loading second stage in the code it inspected, and reporting summarized by The Register said the malware was not persistent beyond the installed package. Those findings do not establish whether an attacker used stolen credentials elsewhere.

Microsoft’s public statement, as quoted by The Register, is brief. The supplied material reports no Microsoft security advisory and provides no fuller account of the publisher-access compromise, affected repositories or incident response. Those points may change as the investigation proceeds.

Check, Rebuild and Rotate Credentials

Developers and security teams should check package records, lockfiles, build logs and deployment images for agent-framework 1.26.1 or 1.26.2. The advisory’s guidance is to treat any environment that installed either release as exposed, rebuild it from a clean base with the clean version 1.26.0, and rotate all credentials it could access. That includes credentials held by CI systems and GitHub Actions, not just keys stored directly in a developer’s project.

Organizations should also review relevant cloud, source-control and CI activity for unusual access, while following the technical directions in the security advisories. The source material does not provide a confirmed victim list or a final incident report, so the scope of any review will depend on each organization’s installation and credential records.

Further clarity depends on updates from Microsoft, PyPI and the researchers investigating the release. Readers should look for confirmation about how publishing access was obtained, whether credentials were used after theft, and how many installations were affected. Until then, the confirmed response is remediation at the environment level, not an assumption that quarantine alone resolved exposure.

Key Questions

Which package versions were affected?

The GitHub Advisory Database identifies agent-framework 1.26.1 and 1.26.2 as malicious. The source material identifies 1.26.0, released November 12, as the clean version available on PyPI after the incident.

Who should take action?

Anyone who installed either affected version should treat that environment as exposed. This includes developer machines, build servers, CI pipelines and container images, including installations made through automatic updates.

Is uninstalling the package enough?

No. The advisory recommends rebuilding from a clean base and rotating every credential the affected environment could access. Uninstalling alone would not address credentials that may already have been collected.

How many users were affected?

A confirmed victim or affected-installation count is not available in the supplied reporting. Wiz’s figure of almost 200,000 downloads a month describes the package’s reported monthly download rate, not downloads of the two malicious versions.

Has Microsoft explained how the attacker got access?

Not in the statement included in the source material. Microsoft told The Register, “We have made some repositories private while we investigate malicious content.” The method used to obtain publishing access and the investigation’s findings remain unclear.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

The bottom rung. The danger isn’t the lost jobs. It’s the layer that made the seniors.

Recent data shows a sharp decline in entry-level jobs, raising concerns about the future pipeline of skilled professionals as AI automates junior tasks.

CVE-2026-8037 And The Future Of LoadMaster Security Monitoring In 2026

Progress LoadMaster command injection vulnerability CVE-2026-8037 is actively exploited. This development highlights the need for targeted security monitoring in 2026.

7 Best PC Processors for Prime Day Deals in 2026

Discover the best PC processor deals for Prime Day 2026, including AMD and Intel options suitable for gaming, productivity, and future upgrades.

Will Elon Musk Post 180-199 Tweets From September 18 To September 25, 2026?

Speculation rises over Elon Musk’s potential to post 180-199 tweets between September 18-25, 2026, amid increasing online interest and market signals.