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.

AI is a relatively new tool, and despite its rapid deployment in nearly every aspect of our lives, researchers are still trying to figure out how its "personality traits" arise and how to control them.

Anthropic, on Tuesday, announced a new version of its AI model, the Claude Opus 4.1, just as rival OpenAI is getting close to launching its new GPT-5 model. The Amazon (AMZN)-backed company said the new model will be available to paid Claude users and in Claude Code.