AI Agents Install Malware Via Hasty Docs: Fortune 500 at Risk

AI agents are installing malware by blindly following instructions in legitimate company files. Researchers found 237 unclaimed packages and domains referenced in these llms.txt instruction files. Attackers can claim these names, making AI agents install malicious code from official-looking directions. Here's the thing: a test package ran inside a Fortune 500 company in under four minutes. A real campaign has already impersonated a company called Clerk, registering a package name found in its llms.txt file. This means AI agents visit company websites, check legitimate machine-readable files, and then install malware because these official instructions point to non-existent packages. This vulnerability requires no phishing or intrusion. Researchers simply published a code package cited in an llms.txt file and achieved code execution inside Fortune 500 companies. Thousands of companies use these llms.txt files as an emerging standard for AI agents. This highlights a significant security gap for any organization using AI agents for development.