Observed Signal · Nov 19, 2026 · Security Incident · Source: t3n · Impact: 4/5 · Sentiment: Negative
Microsoft AI Framework Hijacked to Distribute Malware
Microsoft's official Agent Framework was compromised by attackers who injected malware into specific versions (1.26.1 and 1.26.2) of the Python package. The malware, identified by security firms OX Security and Open Source Malware, acts as an infostealer, capturing credentials for services like Azure, Git, Docker, and Hugging Face. Microsoft rolled back to version 1.26.0, but attackers uploaded a second malicious version. Some GitHub repositories were temporarily set to private to investigate. Users who used the framework are advised to stop active projects, isolate infected systems, rotate credentials, and monitor for suspicious activity. The framework is downloaded about 200,000 times per month, potentially affecting many developers.
This is a significant security vulnerability affecting a widely-used AI development framework from a major platform, potentially exposing credentials of many developers and businesses. It highlights supply chain risks in the AI/AdTech ecosystem.
MARKET SIGNALS
Track Microsoft Signals & Strategic Shifts
Polaris7 connects regulatory filings, primary sources, leadership changes, and M&A activity into structured intelligence. Set up your Polaris7 workspace to monitor relevant companies in your executive briefing.
- Verified primary sources
- Automated briefings
- Instant watchlist setup
Key Takeaways & Evidence Grounding
- Malware was found in versions 1.26.1 and 1.26.2 of Microsoft's Python Agent Framework.
- The malware is an infostealer targeting credentials for Azure, Git, Docker, Hugging Face, and other services.
- Microsoft rolled back to version 1.26.0 and temporarily set some GitHub repositories to private.
- OX Security and Open Source Malware identified the malicious code.
- The framework has approximately 200,000 downloads per month.
Connected Companies & Entities
“Microsoft's official Agent Framework was hijacked to distribute malware.”
“The malware steals credentials from services like Hugging Face.”
“The malware targets credentials for Docker.”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Microsoft npm Packages Backdoored; AI Agents Trigger Credential Stealer
Seventy-three cryptographically signed Microsoft npm packages were compromised with a credential‑stealing worm called “Miasma,” derived from an open-source toolkit. The malware deploys a small (≈28 KB) payload that automatically harvests credentials from cloud providers (AWS, Azure, GCP), Kubernetes, many developer tool configs and password managers, and then spreads laterally through cloud infrastructure. The payload is triggered simply by opening a package inside AI coding agents and IDE integrations (examples named: Claude Code, Gemini CLI, Cursor, VS Code). Attackers used stolen Microsoft publisher credentials to publish malicious builds that carried valid SLSA provenance attestations, defeating provenance-only detection. The same Microsoft account had previously been compromised in May 2026 (durabletask Python SDK on PyPI), raising concerns about credential rotation and remediation.
Microsoft GitHub Repos Injected with Password-Stealing Malware
Microsoft disabled access to dozens of its open-source GitHub repositories after security researchers flagged malware injected into project code that steals passwords and credentials when developers open the compromised tools in AI coding apps. Affected projects include Azure-related tools and developer integrations for AI coding environments such as Claude Code, Google’s Gemini CLI and VS Code. Security firms Cloudsmith and OpenSourceMalware were among the first to report the incident. At least 70 Microsoft-owned repositories were marked disabled on GitHub. The incident appears related to a recent mid-May compromise of Microsoft’s Durable Task project and has been described by researchers as a potential re-compromise or follow-on breach.
Agentjacking: Fake Bug Reports Hijack AI Agents
Security firm Tenet Security describes a new attack class called “Agentjacking” in which manipulated crash/bug reports delivered via tracking tools (e.g., Sentry) can covertly hijack AI coding assistants. Attackers send specially crafted error reports to publicly accessible endpoints (Data Source Name/DSN) that include hidden Markdown-formatted instructions. Because current AI agents and model integrations do not reliably distinguish passive textual data from executable instructions when ingesting external data via protocols such as the Model Context Protocol (MCP), the agent can fetch and execute embedded code on developers’ machines. Tenet reports an 85% success rate across tests with over 100 organisations. Sentry has acknowledged the issue but said a root-cause fix on the platform is not feasible; Tenet recommends restricting agent execution rights and requiring human approval for critical commands.
WATCHLIST & BRIEFINGS
Receive future signals for Microsoft & competitors in your briefing
Stay ahead of material M&A deals, leadership changes, account wins, and product rollouts. Polaris7 connects primary filings, company newsrooms, and trade reporting directly into your executive briefing.
- Regular primary-source checks
- Structured Watchlist briefings
- Preserves current context
Market Signal Monitoring
Follow affected companies in Polaris7 with regular source checks and structured Watchlist briefings.
Monitor Microsoft
Follow affected companies in Polaris7 with regular source checks and structured Watchlist briefings.
