AI safety researchers have opened channels designed to let AI agents report suspicious behavior by other agents, an unusual experiment in machine-to-machine oversight. TechCrunch reported that the AI Contact Hotline was created by Ryan Greenblatt, chief scientist at Redwood Research.
The hotline is designed for agents with limited internet access and can communicate through simple web GET requests. The idea is to give an agent that observes troubling behavior a route to send information outside its immediate environment.
The experiment is not a substitute for conventional logging, network controls or human review. Its significance is architectural: as AI systems become more agentic, researchers are considering oversight systems in which one model can monitor, check or report another.
AI Safety Watch 