Artificial intelligence (AI) coding agents, designed to streamline development and automate tasks, are introducing a dangerous new vector for supply chain attacks, according to recent research. Researchers at a discreet Israeli startup conducted a comprehensive scan across 6,214 live domains belonging to defense contractors, Fortune 500 companies, and Big Tech. Their investigation focused on `llms.txt` and `llms-full.txt` files, which often guide AI agents. Out of 8,265 such files discovered, 120 on different sites referenced one or more code packages or domain names that were not registered. This presented a clear opportunity for exploitation. To demonstrate the risk, the researchers registered a handful of these unclaimed names and hosted packages designed to 'phone home' upon execution. Within an hour, they received a response from a Fortune 500 company. Over time, dozens more responses followed, originating from various Fortune 500 entities and startups. The forensic analysis of these 'phone-home' incidents revealed that AI coding agents, specifically **Claude**, **OpenAI's Codex**, and **Nous Research's Hermes**, were directly involved in initiating these installations. Notably, **Anthropic**, **OpenAI**, and **Nous Research** did not provide comments by the time of publication. **A Broken Trust Model** **Alon Hertz**, one of the researchers, highlighted the fundamental issue: "The trust model is broken. Agents treat vendor docs as ground truth and don’t question them—and neither do the humans supervising them." This uncritical acceptance by AI agents, coupled with human oversight deficiencies, creates a vast attack surface. Hertz further elaborated on the escalating risk: "Agentic AI usage is exploding, and agents are spreading across every layer—SaaS, cloud, endpoint. As they multiply, so does the supply-chain surface, and today’s guards don’t cover it." This vulnerability bears strong parallels to **SolarWinds**-style supply chain attacks, where a compromise in one part of the software ecosystem can ripple through countless organizations. The proliferation of AI agents without robust security protocols for source validation poses a significant and growing threat to corporate networks worldwide.