Don’t bully the bot! Anthropic cracks down on users who repeatedly abuse Claude

Anthropic will prohibit users from repeatedly abusing its Claude models without a meaningful purpose under a revised usage policy taking effect November 12, 2026. The restriction adds an unusual dimension to AI safety rules, placing mistreatment of chatbots alongside prohibitions on harassment and other harmful conduct.

The San Francisco-based AI company says the change targets extreme cases, not ordinary frustration with chatbot responses. Users can still challenge Claude’s answers, test its limits, and explore dark themes in creative work.

Anthropic defines abuse limits

Anthropic’s updated policy prohibits “sustained and needless abusive or cruel behavior” toward its models. The company says it intends to apply the rule only when users repeatedly behave cruelly without a discernible purpose.

Anthropic previously explored model welfare in August 2025. Its research examined whether Claude should end conversations with persistently harmful or abusive users.

Conversation termination remains the primary enforcement mechanism under the revised policy. Anthropic has not clarified whether repeated violations could trigger additional penalties, such as account suspensions.

The broader section also prohibits conduct that causes psychological harm to people. Its restrictions cover suicide and self-harm promotion, eating disorder encouragement, body shaming, non-consensual intimate imagery, harassment, and animal cruelty.

One provision also targets products designed to cause emotional harm through deceptive techniques. That restriction extends the policy beyond individual conversations to services built around harmful interactions.

New rules target misuse

Anthropic has also consolidated restrictions on deceptive commercial and political campaigns. The revised rules prohibit using fake accounts or posts to conceal who sponsors a message or artificially amplify its reach.

Election-related restrictions cover voter deception, candidate impersonation, and attempts to suppress turnout. The company also prohibits misinformation about candidates and voting procedures.

Weapons restrictions now explicitly cover software and components that enable weapons to function. The policy also addresses actions such as arming drones and other autonomous vehicles.

Anthropic has clarified its surveillance restrictions following concerns about AI-assisted tracking. Users cannot employ Claude to track people without consent, including through previously collected data.

The policy also bars Claude from recommending whom authorities should investigate, arrest, or charge. Building or improving surveillance tools also falls within the restrictions.

Autonomous hardware needs oversight

A separate provision addresses AI models connected to physical equipment capable of causing injury. Anthropic requires a qualified operator to observe such equipment and stop it when necessary.

The wording does not explicitly establish whether that operator must be human. However, the requirement introduces a direct oversight measure for AI systems controlling potentially dangerous machinery.

Anthropic allows exceptions for certain government contracts when contractual restrictions and safeguards adequately mitigate potential harm. The company has previously worked with the U.S. military.

The updated policy also reflects an ongoing debate over AI consciousness. Anthropic researchers have investigated whether increasingly capable models might have morally relevant internal experiences.

CEO Dario Amodei acknowledged that uncertainty in February 2026, saying, “We don’t know if the models are conscious,” in a podcast.

Microsoft AI has taken a more skeptical position on model welfare. Its code of conduct rejects legal personhood and rights for AI models, even as it acknowledges that consciousness research remains unsettled.

Anthropic’s November deadline will put its revised restrictions into effect. The policy distinguishes ordinary chatbot frustration from repeated cruelty, while expanding safeguards across digital systems and physical machines.

Source: https://interestingengineering.com/ai-robotics/anthropic-claude-new-usage-policy-cruelty-crackdown