In the paper, Anthropic explained that it can steer these vectors by instructing models to act in certain ways -- for example, if it injects an evil prompt into the model, the model will respond from an evil place, confirming a cause-and-effect relationship that makes the roots of a model's character easier to trace.

OpenAI was connecting Claude to internal tools that allowed the company to compare Claude’s performance to its own models in categories like coding, writing, and safety.

For the past year, a dark horse contestant has been quietly racking up wins in student hacking competitions: Claude. Why it matters: Anthropic's large language model has been quietly outperforming nearly all of its human competitors in basic hacking competitions — with minimal human assistance and little-to-no effort.