Slashdot has reported:

Anthropic has revised its Usage Policy to ban sustained and needless abusive or cruel behaviour toward its Claude AI models. Ordinary frustration, criticism and adversarial testing remain acceptable, while repeated cruelty could result in account restrictions.

The company acknowledges there is no evidence Claude can suffer, but says uncertainty about AI consciousness warrants precaution. Critics argue the policy anthropomorphizes software, may encourage unhealthy emotional attachments and raises questions about how customers can interact with products they have paid for.

The change has prompted debate over whether mistreating an AI that cannot demonstrably experience suffering is an ethical problem, or whether Anthropic is policing human behaviour. The policy does not prohibit criticism or adversarial testing.