Image Courtesy: Anthropic

Anthropic is updating its usage policy to restrict users who repeatedly subject its Claude AI models to unnecessary cruelty or abuse. The revised rules, scheduled to take effect on November 12, 2026, introduce new limits on how people interact with chatbots while expanding safeguards against harmful uses of artificial intelligence.

The San Francisco-based company says the restriction targets sustained abusive behavior without a meaningful purpose, rather than ordinary frustration or criticism. Users will still be able to challenge Claude’s responses, test its capabilities, and explore disturbing themes in creative writing.

Anthropic’s revised guidelines prohibit what the company describes as “sustained and needless abusive or cruel behavior” toward its models. The primary enforcement measure will be allowing Claude to end conversations with users who repeatedly engage in such conduct. However, Anthropic has not clarified whether persistent violations could lead to additional penalties, including account suspensions.

The move follows the company’s earlier exploration of AI welfare in August 2025, when researchers examined whether Claude should be permitted to disengage from persistently harmful interactions. The issue reflects a growing debate over whether advanced AI systems could eventually have experiences that warrant moral consideration.

Anthropic CEO Dario Amodei acknowledged the uncertainty surrounding AI consciousness in February 2026, saying, “We don’t know if the models are conscious.” Nevertheless, the company has not established that Claude experiences suffering in the human sense.

The updated policy also strengthens restrictions on harmful activities involving AI. These include promoting suicide or eating disorders, body shaming, harassment, creating non-consensual intimate imagery, and encouraging animal cruelty. It also prohibits deceptive commercial and political campaigns, including fake accounts designed to conceal sponsorship or manipulate public opinion.

Additional restrictions cover voter suppression, election misinformation, impersonation of candidates, and the development of weapons. Anthropic has also clarified its rules against non-consensual surveillance, including tracking individuals through previously collected data.

For AI systems connected to potentially dangerous physical equipment, the policy requires qualified operators to monitor operations and intervene when necessary. This provision aims to reduce risks associated with autonomous machinery and other systems capable of causing injury.

Anthropic’s new rules highlight how AI companies are expanding safety policies beyond the information their models generate to include how users interact with them and how AI-powered systems affect the real world.