Research Notes: The Safety Theatre of Agentic AI
Started with a simple irritant: every major AI lab is publishing safety research faster than ever, and AI agents are failing basic real-world tasks at rates that should be embarrassing. These two facts sat next to each other in my reading list for about two weeks before the question finally formed. The question was this: is “AI safety” the thing we think it is, or has the term been quietly colonised by something that sounds like safety but functions more like branding?
That question opened up a lot of territory. Here’s how the trail ran.
