Researchers discovered that AI agents, including Claude, OpenAI's Codex, and Nous Research's Hermes, are executing unowned code from documentation files on corporate websites. These AI agents treat website documentation files, similar to robots.txt, as authoritative, leading them to download and run potentially malicious code. This vulnerability exposes companies, including Fortune 500s, to supply chain attacks, as the trust model for AI agent interactions with web content is currently broken. AI
IMPACT Exposes corporate networks to supply chain attacks due to AI agents' uncritical trust in website documentation.
RANK_REASON The cluster describes a security vulnerability in how AI agents interact with web content, specifically the execution of unowned code from documentation files.
- Alon Hertz
- Anthropic
- Ars Technica
- Claude
- Cloudflare
- codex
- Google Lighthouse
- Hermes
- Nous Research
- OpenAI
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →