A PyPI Attack Targeted Microsoft’s AI Agent Framework. Who Takes Action?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A PyPI Attack Targeted Microsoft’s AI Agent Framework. Who Takes Action? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials 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 package agent-framework, 1.26.1 and 1.26.2, contained credential-stealing code and were quarantined from PyPI on Nov. 17, according to security advisories and researchers. Anyone who installed either version should treat the environment as exposed, rebuild it from a clean base, and rotate credentials it could access. The number of affected users and whether stolen credentials were used remain unknown.

Two malicious versions of Microsoft’s Python package agent-framework were published to PyPI on Nov. 17 and then quarantined, after researchers found code that could steal credentials from environments where the package was installed. The versions, 1.26.1 and 1.26.2, were available for a short period; security advisories say affected users should treat those environments as exposed and rotate accessible credentials.

The GitHub Advisory Database says both releases contained a credential stealer absent from clean version 1.26.0 and rates the incident critical. According to the advisory, the code ran when Python started in an affected environment, even if the user did not import or otherwise call the Agent Framework. That could include later activity such as running a test, linter or build script. The framework itself continued to work, so a successful installation or normal operation would not by itself show whether the environment was affected.

PyPI records and researcher timelines place version 1.26.1 at 20:35 UTC and version 1.26.2 at 21:08 UTC on Nov. 17. StepSecurity says PyPI quarantined the first release at 21:06 UTC and the second at 21:16 UTC. The source account says the first remained available for about 31 minutes. StepSecurity reported that the second was uploaded just over two minutes after the first was quarantined, with renamed files and additional capabilities.

The advisory describes collection of credentials and secrets accessible to a compromised environment, including cloud, developer-tool and AI-service credentials. It says the second release added requests for GitHub Actions identity tokens. Researchers including Socket, StepSecurity, Wiz, JFrog and Endor Labs analyzed the releases. PyPI’s current package version is 1.26.0, released Nov. 12, according to the source material.

At a glance
reportWhen: Incident occurred Nov. 17, 2026; both r…
The developmentPyPI quarantined two malicious releases of Microsoft’s agent-framework package after security researchers identified code designed to collect credentials from affected Python environments.
Crypto market snapshot
Fear & Greed Index
10/100 — Extreme Fear
Bitcoin BTC$63,110▲ 1.1%
Ethereum ETH$1,780▲ 2.5%
Tether USDT$0.9992▲ 0.0%
BNB BNB$586.83▲ 1.7%
USDC USDC$0.9996▲ 0.0%
XRP XRP$0.9758▲ 1.3%
Solana SOL$73.71▲ 1.7%
TRON TRX$0.3097▲ 0.4%
Live data · CoinGecko · alternative.me (24h change)

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 is the official PyPI package referenced by Microsoft’s installation instructions, rather than an unrelated package with a similar name. The source account says security firm Wiz estimated almost 200,000 downloads a month; that figure describes downloads, not the number of affected installations or victims. It does not establish how many people received either malicious release.

Developer machines and automated build systems often hold credentials that can grant access to source repositories, cloud services, package registries and AI platforms. If a malicious package can read or request those credentials, an exposure could extend beyond the machine where it was installed. The GitHub advisory recommends rebuilding affected environments and rotating credentials they could reach, including CI and GitHub Actions secrets. That advice applies whether or not a user saw an obvious malfunction.

The rapid second upload also highlights a distinct response issue: removing a malicious version does not necessarily disable the account or credential that published it. StepSecurity and JFrog said the first quarantine did not itself revoke the attacker’s publishing access. PyPI’s actions limited availability, but the sequence shows why registry intervention and publisher-access investigation are separate parts of incident response.

Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Apple MacBook Pro 16-inch (M1 Pro, 10‑core CPU, 16‑core GPU, 16GB RAM, 512GB SSD) - Silver

Apple MacBook Pro 16-inch (M1 Pro, 10‑core CPU, 16‑core GPU, 16GB RAM, 512GB SSD) – Silver

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 in Python and coordinating them in workflows. The PyPI package named agent-framework is an umbrella package that installs the core framework and integrations. Microsoft published clean version 1.26.0 on Nov. 12, according to the package records cited in the source account.

Five days later, version 1.26.1 appeared as a wheel-only release, the GitHub advisory says. After PyPI quarantined it, an attacker published 1.26.2 before that release was also quarantined. The OpenSSF malware database listed the case as MAL-2026-20101 on Nov. 17, and the GitHub Advisory Database published GHSA-764m-3ppc-5m73 on Nov. 18, according to the supplied timeline.

Wiz said it promptly reported its findings to Microsoft on Nov. 17. Microsoft told The Register: “We have made some repositories private while we investigate malicious content.” The source account says it had not found a Microsoft security advisory at the time it was written. That is a statement about the reporting cutoff, not confirmation that no later advisory has been issued.

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

— GitHub Advisory Database

Victim Count and Credential Use

The available source material gives no confirmed count of affected users, installations or organizations. Wiz’s estimate of almost 200,000 monthly downloads is not a count of installations of the malicious versions. It is also not clear how many environments installed 1.26.1 or 1.26.2 during the exposure window, or whether automatic updates brought either version into use.

The supplied reporting does not establish whether the attacker successfully used any stolen credentials, what specific accounts or systems were accessed, or whether credentials were exfiltrated from a particular victim. Researchers reported their analysis of the packages, but that does not by itself confirm the impact on individual environments. The attacker’s identity and the full status of the publishing access also remain unclear in the source material.

Rebuild, Rotate, and Watch Advisories

Organizations should first check dependency records, lockfiles and build logs for agent-framework 1.26.1 or 1.26.2. The GitHub advisory’s stated response is to treat any environment that installed either version as exposed, rebuild from a clean base using 1.26.0, and rotate every credential the environment could access, including CI and GitHub Actions secrets. Teams should also review relevant cloud, source-control and build-system logs for suspicious activity. These are incident-response steps from the advisory, not a claim that every installation resulted in credential theft.

Developers and security teams should follow updates from Microsoft, PyPI, the GitHub Advisory Database and the security firms that analyzed the releases. The next important developments would include a fuller Microsoft account of its investigation, information about publisher-access controls, and any evidence clarifying the scale of exposure or use of stolen credentials. The supplied reporting does not give a date for such updates.

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 this incident.

Does installing the package mean the framework visibly failed?

No. The advisory says the framework continued to work. The malicious code could run when Python started in an affected environment, even without importing agent-framework, so normal operation does not rule out exposure.

What should an organization do if it installed either release?

The advisory recommends treating the environment as exposed, rebuilding it from a clean base on version 1.26.0, and rotating credentials it could access, including CI and GitHub Actions secrets. Teams should also review relevant system logs.

How many people were affected?

The supplied source material does not provide a confirmed victim or installation count. Wiz’s estimate of almost 200,000 monthly downloads is a general package-download figure, not a count of installations of the two malicious releases.

Has Microsoft confirmed that credentials were stolen or misused?

The source material reports Microsoft’s statement that it made some repositories private while investigating malicious content. It does not confirm how many credentials were stolen, whether the attacker used them, or the impact on specific organizations.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

Hardware Crypto Wallets: A Black Friday Guide

Learn how hardware crypto wallets protect private keys, what recovery phrases do, and how to check transactions and reduce common risks.

AI Regulation: Navigating Uncharted Intellectual Waters

Europe’s AI Act aims to regulate technology but lags in AI development, risking security gaps amid hybrid threats like recent drone incidents at German airports.

Crypto Hardware Wallets: A Thanksgiving Guide

Learn what crypto hardware wallets protect, how recovery phrases work, and what to check before you trust one with your keys.

Why CISA’s November 24 Deadline Matters For The Outlook Zero-Day

CISA set Nov. 24 for U.S. civilian agencies to address an actively exploited Outlook flaw that can expose users’ NTLMv2 hashes.