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

🔍 Read the full analysis: Microsoft’s AI Agent Framework, A PyPI Hijack, And The Credential Theft Risk on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Two releases of Microsoft’s Python package agent-framework, versions 1.26.1 and 1.26.2, were found to contain credential-stealing code and were quarantined by PyPI on November 17, 2026. Security advisories say the code could run when Python started in an affected environment; anyone who installed either version should treat accessible credentials as exposed, rebuild from a clean version and rotate them.

Two releases of Microsoft’s Python package agent-framework on the PyPI registry contained credential-stealing code, according to security advisories and researchers who analyzed the files. PyPI quarantined versions 1.26.1 and 1.26.2 on November 17, but anyone who installed either release should treat the environment as exposed and rotate credentials it could access.

The GitHub Advisory Database says the malicious code was absent from the preceding clean release, version 1.26.0, and rates the incident critical. PyPI records and researcher timelines place the first upload at 20:35 UTC on November 17. StepSecurity says PyPI quarantined it at 21:06 UTC. A second release appeared at 21:08 UTC and was quarantined at 21:16 UTC, according to the same account.

The advisory says the added code ran when Python started in an affected environment, without the user importing the framework. It could gather credentials and tokens from the environment and developer tools, and request certain cloud or pipeline identity tokens. Researchers said the package’s ordinary framework functionality continued to work, so installation did not necessarily produce an obvious error.

The advisory recommends treating systems that installed either version as exposed, rebuilding them from a clean base using version 1.26.0, and rotating every credential those systems could reach, including CI and GitHub Actions secrets. The source account says both malicious releases were removed from PyPI and that the registry’s current version was 1.26.0. That status is time-sensitive and should be checked against PyPI before taking action.

At a glance
reportWhen: Incident occurred November 17, 2026; th…
The developmentSecurity researchers reported that two malicious releases of Microsoft’s agent-framework package on PyPI collected credentials before the registry quarantined both versions.

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 Cloud and CI Credentials

The incident matters because software packages often run in environments holding credentials with access beyond a developer’s computer. A compromised laptop, build server, container or CI worker could expose cloud keys, repository tokens and AI service credentials, depending on what it could access. Stolen tokens may allow access to company systems or cloud resources; the available source material does not establish that such access or misuse occurred.

The reported ability to run at Python startup broadens the exposure beyond a direct use of the affected framework. A build step, test or other Python process could trigger the code after installation. Security teams therefore need to check package installation records and the credentials available to affected environments, rather than relying only on whether an application imported agent-framework.

The short interval between the two uploads also highlights a registry-response limitation described by researchers: quarantining a release does not necessarily revoke a compromised publisher credential. StepSecurity and JFrog both said that the first quarantine did not, by itself, stop a further upload. PyPI’s actions removed the listed releases, but the source does not establish how the attacker obtained publishing access or whether that access was revoked.

Products Worth Considering

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Apple MacBook Air 13-inch with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and 12MP Center Stage Camera – Midnight

Apple MacBook Air 13-inch with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and 12MP Center Stage Camera – Midnight

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 described in the source material as an umbrella package that installs the framework’s core code and integrations. Microsoft’s documentation directs users to the official package, making the incident different from a similarly named copycat package.

PyPI records show that clean version 1.26.0 was published November 12. The GitHub Advisory Database and security researchers then documented malicious versions 1.26.1 and 1.26.2 on November 17. StepSecurity reported that the second upload came just over two minutes after the first was quarantined. Researchers at Socket, StepSecurity, Wiz, JFrog and Endor Labs were among those who analyzed or reported the releases, according to the source account.

Wiz said it reported its findings to Microsoft on November 17. The source material also reports that Microsoft gave The Register a short statement: “We have made some repositories private while we investigate malicious content.” It says no Microsoft security advisory had been found at the time of publication. The independently published GitHub and OpenSSF advisories provide the public guidance described here.

““a credential stealer that is not present in the last clean release, 1.26.0””

— GitHub Advisory Database

Victim Count and Access Remain Unknown

The source material gives no confirmed count of affected installations or victims, and it does not say how many credentials, if any, were successfully collected or misused. PyPI’s download figures are not presented as a count of installations during the malicious window, so they cannot establish the number of affected systems.

The route by which the attacker gained publishing access is also unclear. The source reports that researchers found no remote-code-loading second stage in the inspected code and that the malware was not persistent beyond the installed package. Those findings do not establish whether stolen credentials were used elsewhere, whether every affected file was identified, or whether other systems or repositories were involved. The extent of Microsoft’s investigation and any further official guidance were not available in the supplied material.

Verify Exposure and Rotate Secrets

Organizations and developers should check dependency locks, package installation logs, build records and container histories for versions 1.26.1 or 1.26.2. If either was installed, the advisory’s guidance is to rebuild from a clean base on the clean release and rotate credentials that the environment could reach. Teams should include credentials held by CI systems and automation, not just keys stored in a developer’s project files.

Security teams should follow the GitHub Advisory Database, OpenSSF’s malware listing, PyPI’s package record and Microsoft’s communications for changes to the incident status or remediation advice. The source material does not provide a date for a final Microsoft report or say when investigators will publish a victim count. Until those details are confirmed, affected organizations should preserve relevant logs and treat accessible credentials as potentially exposed.

Key Questions

Which package versions were affected?

The reported malicious PyPI releases were agent-framework 1.26.1 and 1.26.2. The advisory identifies 1.26.0 as the preceding clean release.

Could the code run without using the Agent Framework?

Yes. The GitHub Advisory Database says the code ran when Python started in an affected environment, without importing agent_framework.

What should someone do after installing an affected version?

Treat the environment as exposed, rebuild it from a clean base using the clean release, and rotate credentials it could access. The advisory specifically includes CI and GitHub Actions secrets in its guidance.

How many systems or credentials were affected?

The source material reports no confirmed victim or credential count. PyPI download figures do not establish how many systems installed the malicious versions during the exposure window.

Did Microsoft publish a full response?

The source material quotes Microsoft as saying it had made some repositories private while investigating malicious content. It reports no Microsoft security advisory at the time, so further details and the investigation’s outcome remain unclear.

Source: ThorstenMeyerAI.com

You May Also Like

How to Pick the Right Tripod for a Talking-Head Studio

Just selecting a tripod involves more than stability—discover how to choose the perfect one for your talking-head studio needs.

The Monitor Upgrade That Makes Analytics Easier to Read Fast

An upgraded monitor with accurate color and high refresh rate enhances data analysis clarity, making complex insights easier to spot—discover how to optimize your setup.

Why Smart Locks and Biometric Safes Belong in Some Creator Studios

Just imagine how smart locks and biometric safes could transform your creator studio’s security—discover why they’re essential for safeguarding your valuable assets.

How to Build a Home Office Network That Can Actually Scale

Discover how to build a scalable home office network that adapts to your growing needs and keeps your work and personal devices secure.