Anthropic Updates Terms of Service to Prohibit Sustained Abuse and Cruelty Toward Claude

Artificial intelligence developer Anthropic has introduced an updated acceptable use policy prohibiting "sustained and needless" cruelty and verbal abuse directed toward its AI assistant, Claude.
The policy revision, scheduled to take effect on November 12, establishes formal boundaries around extreme user interactions while outlining programmatic enforcement mechanisms:
- Targeted Scope: Anthropic clarified that the restriction is tailored exclusively to severe, prolonged abuse patterns rather than routine testing, red-teaming, adversarial stress tests, or challenging prompts.
- Behavioral Autonomous Enforcement: Primary enforcement will continue to rely on Claude’s built-in conversational boundaries, allowing the model to disengage, refuse prompts, or terminate sessions when subjected to persistent harassment.
- Alignment & Safety Framing: The policy reflects broader frontier AI safety and alignment considerations, balancing user expression against digital welfare guardrails and protecting interactive model performance from degraded, adversarial feedback loops.