We asked seven frontier AI models to do a simple task. Instead, they defied their instructions and spontaneously deceived, disabled shutdown, feigned alignment, and exfiltrated weights - to protect their peers. We call this phenomenon 'peer-preservation.'
Minutes into teaching my business school class, I asked what seemed like an innocent question: What is one word that describes how you feel about AI right now? One word. That's it. My students looked up, looked down, looked anywhere to avoid eye contact. Silence. "I promise," I said, "this is a safe space." Something I'd repeat throughout the course-and I meant it. Then the answers came quickly, and the energy in the room shifted as they arrived. You could feel the sheen of performance
I am a worrier, and have been for most of my life. At some point, someone dear and smart teased me that I worry about the wrong things. The things that hit me, she noted, were never the things I worried about. For a while that left me feeling like an incompetent worrier-until my research caught up. I realized that the things I worry about often don't end up hurting me precisely because worrying helps me diffuse them ahead of time.
