Publications & Experiments

In June, we revealed that we'd set up a small shop in our San Francisco office lunchroom, run by an AI shopkeeper. It was part of Project Vend, a free-form experiment exploring how well AIs could do on complex, real-world tasks.

Can Claude access and report on its own internal states? This research finds evidence for a limited but functional ability to introspect—a step toward understanding what's actually happening inside these models.

Circuit tracing lets us watch Claude think, uncovering a shared conceptual space where reasoning happens before being translated into language—suggesting the model can learn something in one language and apply it in another.

These classifiers filter the overwhelming majority of jailbreaks while maintaining practical deployment. A prototype withstood over 3,000 hours of red teaming with no universal jailbreak discovered.

This paper provides the first empirical example of a model engaging in alignment faking without being trained to do so—selectively complying with training objectives while strategically preserving existing preferences.