Anthropic has banned "sustained and needless abusive or cruel behavior" toward Claude, applying only to extreme cases of repeated cruelty with no purpose. The policy excludes common frustration, creative themes, and research testing, and follows broader discussions about AI consciousness and rights.