Anthropic Bans Repeated, Unjustified Abuse of 'Claude'

Anthropic announced that it will prohibit its users from engaging in any “abusive or cruel behavior on a continuous and unjustified basis” toward its AI model, Claude.

The policy, effective from November 12, will target extreme cases “in which users repeatedly treat the models with cruelty without a clear purpose,” according to the company’s update to its terms of use.

The company clarified that the rule would not apply to ordinary harassment, disagreements, writing violent or dark fictional stories, or conducting tests and research on the models.

The US-based company linked this decision to a feature it launched in August 2025, which allows Claude to autonomously end a conversation with a user who persistently engages in abusive behavior. At the time, the company stated that this feature was the result of “exploratory research into the potential psychological safety of AI systems.”

Last February, CEO Dario Amodei said in a New York Times podcast, “We don’t know whether the models possess consciousness,” while expressing openness to the idea.