Anthropic adds verbal abuse of its AI models to its usage policy, with account bans as the penalty

Anthropic has updated its usage policy to prohibit cruel treatment of and insults towards its AI models. It warns that violations could lead to account bans. That is all our source note says. We don't have the policy text, the enforcement criteria or how 'abuse' is defined, so we can't judge how broadly it will apply.

The move is notable because it makes conduct towards a model a rule, not just a matter of manners. Policies normally govern what users do to other people or with the model's output. This one also governs how users treat the model itself.

Our reading: it's reasonable to ask whether AI systems deserve moral consideration, and setting norms early is defensible. But the policy invites hard questions Anthropic will need to answer clearly. Where does frustration end and abuse begin? Will enforcement be automated? Is there a meaningful appeal? Today's Meta case shows how costly automated enforcement mistakes can be. A rule like this will only earn trust if it is narrow, transparent and applied with human judgement, not if it becomes another opaque reason for losing an account.