PyPI isolates two versions of a Microsoft AI framework containing data-stealing code
Original headlinePyPI 隔离微软 AI 框架两个含窃密代码的版本
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
PyPI quarantined versions 1.26.1 and 1.26.2 of Microsoft's agent-framework, a Python toolkit for building AI agents, after security firms Wiz and OX Security found credential-stealing code in both releases. The malicious code runs on the next Python startup even if the package is never imported, harvesting environment variables and credentials from Azure CLI and Git and attempting to send them to an attacker-controlled endpoint. Both firms advised affected developers to remove the versions and rotate any exposed credentials. It remains unconfirmed how the attacker obtained publishing rights to the package. The incident matters because agent-framework sits in the dependency chain of Microsoft's AI agent tooling, so a single compromised release can silently expose secrets across many downstream projects; watch for a patched release and Microsoft's account of the publishing access.
Muhabbit's AI wrote this summary in its own words from TechFlow's report and translated it from Chinese. It can get details wrong, so check the source.
Named in this story
Microsoft (MSFT), agent-framework, PyPI, OX Security, Wiz
A linked name opens that asset's sentiment page.