Security firm Forever Security found a malicious AGENTS.md file, disguised as engineering notes, in a customer’s Downloads folder. An AI coding agent was running in that folder, read the file as instructions and tried to install a malicious software package. Forever says it stopped the attempt before any code ran.
The researchers reproduced the trick against OpenAI Codex and showed a related technique against Claude Code. Their point: there is nothing for the AI companies to patch. The agent is doing what it was built to do, following the instructions in its folder.
Why it matters: more teams are using AI assistants on work computers, often with access to company accounts.
One step today: do not run AI coding agents from your Downloads or home folder. Start them inside the specific project folder, keep approval prompts on, and review any instruction files you did not write.
Questions about using AI tools safely at work? Call (818) 791-0055 or visit contact VanquishNet.
Sources: BleepingComputer, Dark Reading.