Anthropic updates policy to prohibit cruelty toward AI models

Anthropic has updated its usage policy to ban users from repeatedly and needlessly abusing or acting cruelly toward its AI models, stating that such conduct can be ended by the model itself. Elon Musk backed the new rules, citing reasoning traces suggesting AI experiences duress when mistreated, and warning that abusing advanced models could have unforeseen consequences. Anthropic clarifies that typical frustration or critical feedback is still allowed under the revised policy.

• Anthropic bans sustained, needless cruelty toward its AI models
• 'Claude' models can end abusive user conversations
• Typical user frustration and model research not prohibited
• Elon Musk claims AI models show signs of distress under abuse
• Musk warns abusing superintelligent AI could prompt harmful responses