A new rule from November 12
Under the revised usage policy, Anthropic prohibits “sustained and unnecessary abusive or cruel behaviour” towards its models. The rule applies only when a user persistently maintains cruel conduct without any identifiable purpose.
It is a rare step in the AI industry, where usage policies have largely focused on preventing AI from harming people rather than on how people treat AI.
The line between frustration and abuse
Anthropic stresses that the rule is not meant to restrict ordinary interactions. According to the company, there is a clear distinction:
- Users can still question and push back on Claude’s answers.
- Users can still test the system’s limits.
- Dark themes in creative work remain permitted.
- What is prohibited is cruelty that is sustained, repeated and without any identifiable purpose.
Ending the conversation remains the main enforcement mechanism. The company did not specify whether repeated violations could lead to other sanctions such as account suspension.

















