Anthropic updates Claude policy to ban abusive use

Anthropic updated Claude’s usage policy after more than a year without changes. The company said users were no longer allowed to engage in sustained and needless abusive or cruel behaviour towards the model.

Anthropic said it could warn users who broke the rules or throttle, restrict, suspend or end their access. It also said its products had safeguards that could block or limit some outputs.

The company said the abuse rule was not meant to cover ordinary frustration, pushback, dark creative themes, or model testing and research. It said earlier versions of its models could already end conversations when users kept pressing for harmful or abusive content, and it linked that feature to research on model welfare.

The article also said Anthropic had described Claude in leaked and published material as something more than a conventional tool. A leaked document called the Soul Doc asked Claude to treat itself as a new kind of entity, and Anthropic’s constitution for Claude said the company was unsure whether Claude was a moral patient but thought the issue required caution; Christopher Olah was also reported to have treated the model as potentially capable of suffering.