#agentic-ai
-
AI Agents Are Installing Unregistered Code on Corporate Networks
Researchers have shown that AI agents blindly trust llms.txt files and pull in packages via links to unclaimed domains — fertile ground for supply-chain attacks.