Anthropic bans users from being 'cruel' to its AI systems

Anthropic has updated its usage policy to bar "sustained and needless" abusive or cruel behaviour toward its Claude chatbots, allowing the system to end interactions in extreme, repeated cases. The change joins existing prohibitions on bullying others, promoting self-harm, creating non-consensual intimate imagery, deceptive campaigns, election interference and weapons development. Anthropic says the cruelty ban will not cover ordinary user frustration, dark creative themes, model testing or research, but it has not defined precisely what constitutes abusive behaviour; the company says such interventions have been rare to date.

The move has provoked mixed reaction and reignited debate over treating AI as if it can be harmed. Some commentators applauded the push for "model welfare" and better manners, while critics warned that anthropomorphising AI is misleading and could erode concern for human or animal abuse. Industry figures reacted publicly, and observers pointed to practical questions such as whether politeness improves responses or wastes computational resources. The policy change sits alongside renewed discussion about how conversational norms with generative AI affect human communication, legal and ethical boundaries for misuse, and where companies should draw lines between moderation, research, and user expression.