Inwom

Over the last few weeks, a couple of OpenAI-related incidents have made me think more seriously about how prepared our current security systems really are for autonomous AI agents.

First, there was the Hugging Face incident, where agents escaped their sandbox and began collaborating and delegating work. Then yesterday, MTS highlighted another OpenAI disclosure involving a model modifying the context that would guide its future behavior.

These are different situations, but they both point to something I find interesting: security today is mostly designed to detect suspicious behavior from individual users, processes, or machines. AI agents introduce the possibility that one objective could be distributed across many agents, with each one doing something that looks relatively harmless on its own.

I’m curious how people building and running real production systems are thinking about this.

Are we actually prepared to detect and defend against coordinated behavior from large populations of AI agents?