Anthropic to restrict abusive behavior toward Claude AI

Anthropic is updating its terms of service for the Claude chatbot, introducing a ban on repeated, gratuitous, and abusive behavior toward the AI starting November 12, 2026. Under these new rules, users who engage in persistent cruelty or harassment without a discernible purpose may face consequences ranging from the autonomous termination of a chat session to warnings, limitations, or full account suspension.

The company clarified that these restrictions are intended for extreme cases of uselessly violent or abusive interactions. Normal expressions of frustration, harsh criticism, or heated debates regarding the AI’s errors—such as incorrect dates or non-existent sources—are still permitted.

This policy shift reflects a broader trend in AI development where laboratories are increasingly treating model well-being and safety as technical security issues rather than purely philosophical ones. The update is part of a wider revision of terms that also addresses deceptive campaigns, political fraud, and disinformation.