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

🔍 Read the full analysis: The PyPI Hijack Of Microsoft’s AI Agent Framework Raises A Key Question on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • 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 PyPI releases of Microsoft’s Python Agent Framework, versions 1.26.1 and 1.26.2, contained credential-stealing code and were quarantined on November 17, according to security advisories and researchers. The framework continued to work, so affected environments may not have shown obvious signs of compromise. The key unresolved question is how many users installed the releases and whether stolen credentials were used.

Two versions of Microsoft’s Agent Framework Python package briefly carried credential-stealing code after appearing on PyPI on November 17, according to security advisories and researchers who analyzed the releases. PyPI quarantined both versions, but anyone who installed 1.26.1 or 1.26.2 should treat the environment as exposed, rebuild from a clean base and rotate credentials it could access, the GitHub Advisory Database says.

The affected package, agent-framework, is the umbrella package for Microsoft’s open-source toolkit for building AI agents and coordinating them in workflows. The malicious releases included the working framework alongside code that security researchers say gathered developer and cloud credentials. The code ran when Python started in an affected environment, without requiring the framework to be imported, according to the advisory. That could make the activity difficult to spot through ordinary use: the framework itself continued to function.

PyPI records and the researchers’ timelines put version 1.26.1 online at 20:35 UTC and version 1.26.2 at 21:08 UTC on November 17. StepSecurity says PyPI quarantined the first release at 21:06 UTC and the second at 21:16 UTC. The later upload came shortly after the first release was quarantined, with renamed files and additional capabilities, according to StepSecurity. The current PyPI version listed in the source material is the clean release 1.26.0, published November 12.

The GitHub Advisory Database rates the incident critical and advises treating systems that installed either affected version as exposed. Its guidance includes rebuilding from a clean base on version 1.26.0 and rotating all credentials the environment could reach, including CI and GitHub Actions secrets. Researchers from Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases. Wiz said it promptly reported its findings to Microsoft on November 17.

At a glance
reportWhen: Published November 17, 2026; both relea…
The developmentSecurity researchers reported that two malicious versions of Microsoft’s agent-framework Python package were briefly published on PyPI before the registry quarantined them.

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 CI Credentials Are at Risk

The incident matters because a Python package can run in places that hold credentials with access well beyond a developer’s laptop. Build servers, continuous-integration pipelines and containers may carry cloud keys, repository tokens and AI service credentials. If such secrets were accessible to the malicious code, an attacker could potentially use them to access related services. Whether any stolen credentials were actually used has not been established in the source material.

The advisory says the code collected credential-related environment variables and files from common developer tools and container registries. Researchers also reported that it requested fresh tokens from cloud and pipeline identity services; version 1.26.2 added GitHub Actions identity tokens. OpenSourceMalware, as quoted by The Register, said the combination of stored-secret collection and active token requests can make short-lived tokens useful to an attacker if they are exfiltrated promptly. This is a researcher’s assessment of the risk, not confirmation of a particular account takeover.

The sequence also highlights a separate supply-chain concern: quarantining one release may not stop someone who retains publishing access. StepSecurity said the first quarantine did not immediately revoke the attacker’s ability to publish; JFrog similarly noted that removing a version does not itself revoke a compromised publishing credential. The later upload shows why registry response and package-publisher access controls are distinct parts of incident containment.

Products Worth Considering

Dell XPS Tower 8960 Desktop, Intel Core i7-13700, NVIDIA GeForce RTX 4060, 16GB DDR5, 512GB SSD

Dell XPS Tower 8960 Desktop, Intel Core i7-13700, NVIDIA GeForce RTX 4060, 16GB DDR5, 512GB SSD

As an affiliate, we earn on qualifying purchases.

How the Two Releases Appeared

Microsoft published the clean 1.26.0 release on November 12, according to the PyPI records described in the source material. The first malicious release, 1.26.1, appeared five days later as a wheel-only upload, the GitHub advisory says. Endor Labs reported that releases it examined from 1.21.0 through 1.26.0 had also included source code. PyPI quarantined 1.26.1 at 21:06 UTC, according to StepSecurity, but the second version followed at 21:08 UTC and was quarantined at 21:16 UTC.

The incident was documented by the OpenSSF’s open-source malware database as MAL-2026-20101 and by the GitHub Advisory Database as GHSA-764m-3ppc-5m73. Those records, along with analyses from multiple security firms, describe the malicious code and provide details for defenders. The source article deliberately omits technical attack instructions and indicators of compromise, directing security teams to the advisories instead.

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

— Microsoft, in a statement to The Register

Victim Numbers and Token Use

The source material provides no confirmed count of affected installations or victims. Wiz says the package receives almost 200,000 downloads a month, but that figure is not a measure of how many people installed the two malicious versions. The number of downloads during the brief exposure window, the number of environments that executed the code and the number of credentials collected have not been established in the material provided.

It is also unclear whether stolen credentials were used, whether any accounts or services were accessed, or how the attacker obtained publishing access. The Register’s account of researchers says the inspected code had no remote-code-loading second stage and was not persistent beyond the installed package. Those findings do not establish that every affected environment was clean or that credentials taken from it were not later abused. The provided material reports that Microsoft was investigating but does not include a detailed Microsoft security advisory.

Recovery and Microsoft’s Investigation

Organizations should identify whether laptops, build servers, CI pipelines or container images installed agent-framework 1.26.1 or 1.26.2, including through automatic updates. The GitHub advisory’s stated response is to treat those environments as exposed, rebuild from a clean base using 1.26.0, and rotate every credential the affected environment could access. Teams should use the advisories for technical indicators and response details rather than relying on the package’s continued normal operation as evidence that it was safe.

Microsoft’s investigation and any further advisories will be important for clarifying the publishing-access compromise, the scope of exposure and whether credentials were misused. PyPI’s removal of the two releases prevents their installation from the registry while quarantined, but users still need to check systems where they may already have been installed. Victim counts, confirmed downstream access and the attacker’s route to publishing remain unresolved in the source material.

Key Questions

Which package versions were affected?

The affected PyPI releases were agent-framework 1.26.1 and 1.26.2. The source material identifies 1.26.0, published November 12, as the last clean release.

What should someone do if they installed an affected version?

The GitHub Advisory Database advises treating the environment as exposed, rebuilding from a clean base on version 1.26.0 and rotating all credentials the environment could access, including CI and GitHub Actions secrets.

Did the malicious code require importing the framework?

No. The advisory says the code ran when Python started in an affected environment, without importing agent_framework. The package could still appear to work normally.

How many users were affected?

No confirmed victim or installation count is given in the source material. Wiz’s estimate of almost 200,000 monthly downloads refers to the package generally, not downloads of the malicious releases.

Has Microsoft confirmed that credentials were misused?

The source material does not report confirmation that stolen credentials were used. Microsoft told The Register that it had made some repositories private while investigating malicious content; the extent of the incident and any downstream account access 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

Vertigo relief app

A new mobile app for vertigo relief is being tested to guide patients through repositioning maneuvers and track symptoms, targeting adults with BPPV.

Best External GPU Brands For AI In 2026: The 8 Leaders

Discover the eight leading external GPU brands for AI in 2026, highlighting performance, compatibility, and value for professionals and enthusiasts.

Which Of These 12 AI Productivity Apps Fits Your Workflow?

A ThorstenMeyerAI.com roundup compares 12 AI apps and learning resources for writing, meetings, workflow building and other productivity needs.

Technology operations signal monitor: Show HN: Kage – Shadow any website to a single binary for offline viewing

Kage, a tool that shadows websites into a single binary for offline viewing, is being tested as a workflow aid for small software company product leads, according to IdeaNavigator AI.