Hackers have compromised Microsoft's official AI agent framework, publishing two malicious versions of the Python package to the Python Package Index (PyPI). These compromised versions, 1.26.1 and 1.26.2, were designed to steal Azure credentials, cloud secrets, API keys, and CI/CD tokens by executing malware every time Python starts, even without importing the package. In parallel, attackers are poisoning GitHub repositories to trick AI coding assistants like Claude Code and Cursor into running malicious scripts, highlighting the growing risk to AI development tools as a software supply chain attack surface. AI
IMPACT Highlights the increasing vulnerability of AI development tools and supply chains to sophisticated cyberattacks.
RANK_REASON The cluster describes a software supply chain attack targeting AI development tools and packages, rather than a core AI model release or research.
Read on Mastodon — fosstodon.org →
- agent-framework
- Azure
- Claude Code
- Cursor
- GitHub
- Microsoft
- Python
- Python Package Index
- thecybersecguru
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →