Anthropic updates usage policies to prohibit extreme AI abuse

Anthropic is implementing new usage policies, effective November 12, 2026, that prohibit sustained and unnecessary abusive or cruel behavior toward its Claude AI models. The company stated the measure is intended for extreme cases where users act cruelly without a discernible purpose.

To enforce these rules, Claude may terminate specific conversations on Claude.ai and Claude Code. However, the policy explicitly excludes common user frustration, critical feedback, dark creative themes, and model testing or research. Anthropic clarified that these changes do not imply that Claude possesses consciousness or feelings, though the company noted it is exploring the uncertain moral status and potential welfare of AI models.

Beyond behavioral restrictions, Anthropic has also introduced technical safeguards to combat “distillation attacks”—attempts to copy model capabilities into smaller systems—reporting a 90–99% reduction in successful data collection. Additionally, the company updated its terms to clarify prohibitions regarding the use of AI for deceptive activities, weapon software development, and unauthorized surveillance.