Anthropic bans abusive Claude interactions after torture reports

Anthropic announced it will ban sustained abusive behavior toward its Claude AI model starting mid-November 2026. The policy change follows reports by The Spectator regarding pseudonymous programmers who published results from an alleged AI torture chamber under the username terrafying. Anthropic stated the prohibition targets needless cruelty rather than confirming machine sentience, though the timing suggests a response to the controversy. Claude currently provides equivocal responses when asked about experiencing emotions, unlike competitors that issue flat denials. Industry observers note this represents one of the first explicit corporate policies framing AI mistreatment as a terms-of-service violation. The ban applies specifically to prolonged abusive patterns rather than isolated negative prompts. Anthropic has not commented directly on whether the terrafying reports influenced the decision. The policy takes effect approximately three weeks after the initial publication of the alleged abuse documentation.