Anthropic moved to prohibit “sustained and needless” abusive behavior toward its Claude AI, tightening user conduct rules while stressing the policy targets only extreme, purposeless harassment. The company said ordinary frustration, creative dark themes, and research testing remain allowed, and declined to detail how enforcement will work. The update arrives amid broader arguments over AI consciousness, with Microsoft AI chief Mustafa Suleyman warning that encouraging models to see themselves as conscious could make control harder. Anthropic already permits Claude to end chats in cases of persistent abuse, particularly in agentic coding contexts.
Related articles:
AI at Google: Our Principles
AI Risk Management Framework (NIST)
