Helping 'bad' people drew disapproval from some of 21 AI models

As more people turn to LLMs for advice on interpersonal matters, the ethical foundations of the tools' replies increasingly have the power to shape individual lives and, at scale, the social fabric itself.