Anthropic's new rule: Continuously "abusing" Claude also counts as a violation.

Beating AI News Flash: Anthropic has updated its usage policy, explicitly prohibiting users from persistently and without justification abusing or being cruel to Claude and its other AI models for the first time. The new rules will take effect on November 12.

Anthropic emphasized that the ban only targets extreme cases of repeatedly mistreating the model maliciously without a clear purpose. Users who get angry because of incorrect answers, criticize the model, or conduct normal testing and research will not be affected.

In fact, as early as last August, Anthropic allowed Claude to proactively end extremely malicious conversations. This time, it is formally classifying such behavior as a violation. The company said the main approach is still to have Claude end the current chat, and users can start a new conversation. As for whether account bans will be added as a penalty, Anthropic has not yet stated.

This is related to Anthropic's stance on AI consciousness. The company has been researching whether models may have subjective experience, but has not yet reached a definitive conclusion. The new rules sparked debate on Reddit. Some support them, arguing that AI should not be treated as an outlet for venting emotions; others question whether Anthropic is treating AI too much like a person.

The same update also clarified bans on fake account promotion, weapons control software, and monitoring others without consent.