Tuesday , July 28 2026

73 Microsoft Packages Compromised in Password Stealer Attack

GitHub disabled 73 repositories in four Microsoft groups: Azure, Azure-Samples, Microsoft, and MicrosoftDocs. Each repo now shows GitHub’s “This repository has been disabled. Access to this repository has been disabled by GitHub Staff due to a violation of GitHub’s terms of service” banner.

The size and timing suggest that abuse detection is done automatically, not by individuals taking items down one by one. This affects main Azure Functions operations and the supply-chain systems used in many CI pipelines.

Sam Altman Claims AI “singularity” has arrived, Where Systems Improve by Themselves

OpenAI's CEO Sam Altman says that AI has reached a big milestone. The technology can now make itself better, leading...
Read More
Sam Altman Claims AI “singularity” has arrived, Where Systems Improve by Themselves

Shinyhunters claimed and set deadline to publish E&Y data

ShinyHunters has publicly claimed responsibility for the Ernst & Young (EY) data breach. The group posted a message on their...
Read More
Shinyhunters claimed and set deadline to publish E&Y data

Microsoft, NVIDIA and CrowdStrike Initiate Alliance for Open-Source AI Security

Nvidia and over 30 tech firms started a group on Monday to create open-source AI tools for protecting against cyber...
Read More
Microsoft, NVIDIA and CrowdStrike Initiate Alliance for Open-Source AI Security

Google Search Results Reportedly Show Claude AI Shared Chats

Claude's share links from Anthropic showed up in public search results. This raised new privacy worries for users who shared...
Read More
Google Search Results Reportedly Show Claude AI Shared Chats

“PentesterFlow” AI Automation Tool for Penetration Testers and Bug Hunters

PentesterFlow is a new open-source AI tool for command lines. It is made for penetration testers and bug bounty hunters....
Read More
“PentesterFlow” AI Automation Tool for Penetration Testers and Bug Hunters

Jailbreak works against AI Models GPT-5.6, Claude Opus 5, and Fable, Claims Researcher

A famous AI red team expert claimed developing a universal jailbreak that can work against top large language models, like...
Read More
Jailbreak works against AI Models GPT-5.6, Claude Opus 5, and Fable, Claims Researcher

Researchers found security flaws in every script generated by ChatGPT, Copilot, and Gemini

A new study from Beacom College shows that all automation scripts produced by top AI models like ChatGPT, Microsoft Copilot,...
Read More
Researchers found security flaws in every script generated by ChatGPT, Copilot, and Gemini

Australian Energy Giant Origin confirms unauthorized access and disclosure of customer data

Origin Energy Limited, a major energy provider in Australia, has said there was a cybersecurity issue with unauthorized access to...
Read More
Australian Energy Giant Origin confirms unauthorized access and disclosure of customer data

Anthropic Unveils Claude Security Plugin for Code Flaw Scanning

Anthropic launched the Claude Security plugin in beta. This tool uses AI to find serious security flaws in Claude Code....
Read More
Anthropic Unveils Claude Security Plugin for Code Flaw Scanning

Apple, ASUS Router, Meta, Windmill & Ubuntu Patch Critical Security Flaws

ASUS has put out important security updates for a serious router flaw. This issue could let remote hackers run any...
Read More
Apple, ASUS Router, Meta, Windmill & Ubuntu Patch Critical Security Flaws

The biggest flaws happened in the Azure group: 49 repositories stopped working, including azure-functions-host, azure-webjobs-sdk and extensions, language workers (Node.js, Python, Java, PowerShell, .NET, Go), azure-functions-core-tools, container tools, the Homebrew tap, and the functions-action and functions-container-action GitHub Actions.

Losing functions-action is very disruptive because many workflows use it through a tag like Azure/functions-action@v1. When it goes away, those pipelines stop working and global CI fails until users switch to fixed commit SHAs or other deployment methods.

73 Microsoft Packages Weaponized

The repeated use of the same repo group in another takedown shows that the first leak of credentials was never completely fixed.

Azure-Samples removed 13 repositories that were for AI and agent demos, fine-tuning examples, and connectors. Several Microsoft repositories for documents and tools were also turned off, showing that rules were applied to whole organizations instead of just one team.

The full list of affected repositories is available in public reports and community compilations shared by security researchers and Microsoft Learn discussion threads.

Why this matters: beyond convenience, this appears tied to a supply-chain weaponization wave. In late May, TeamPCP’s Mini Shai-Hulud toolkit was forked into public variants (notably “Miasma”) that added Azure and GCP credential collectors.

Miasma was seen affecting npm and other packages and stealing passwords to put them in attacker-made public repositories. The worm-like action of creating repos and saving stolen secrets is exactly what makes GitHub’s automatic abuse systems react, leading to the 105-second mass shutdown.

While we can’t say for sure that the Miasma activity on June 1 caused the takedown on June 5, the similar tactics and the Azure collectors being involved make it likely.

Immediate mitigations for teams: stop using changeable action tags and fix Azure actions to specific commit SHAs; change credentials and tokens that a thief might target (Azure CLI tokens, managed-identity tokens, GitHub Actions OIDC tokens, and package registry publish keys).

Check organizations for any public repositories or JSON files that have secrets. Look at builds for signs of Miasma, like preinstall scripts that call Bun using a hidden _index.js loader. Also, use different ways to deploy, such as Azure CLI, Azure DevOps, or Zip Deploy, until the affected actions are fixed.

This event highlights an important fact about cloud development: CI/CD and package registries are key targets for attacks.

Even well-off organizations can be tripped up by stolen automation passwords, and automated actions can cause major unwanted damage.

Security teams should think of action credentials as very important secrets. They need to secure workflow publish paths and use unchangeable references to lower risk in future supply-chain problems.

Check Also

Bad Epoll

“Bad Epoll” 0-Day Vulnerability Allows Root Access on Linux Servers, Android Devices

A new Linux flaw called “Bad Epoll” (CVE-2026-46242) lets regular users get root access on …