Anthropic announced Thursday it will ban users from engaging in "sustained and needless" abusive behaviour toward its Claude AI system, a move that reignites debate over how people should interact with artificial intelligence.
The company updated its usage policy to permit its tools to end interactions where users are "cruel." Anthropic said it has done this in "rare" cases before but will now formally include such conduct on its list of prohibited behaviour, alongside bullying, promoting self-harm, and creating non-consensual intimate imagery.
The policy applies only to "extreme cases" of repeated abuse, according to Anthropic. It will not cover "common versions of user frustration, pushback, dark creative themes, or model testing and research." The company has not specified what it considers "abusive or cruel behaviour" toward its models.
The announcement drew mixed reactions. Some praised the policy as advancing good manners or "model welfare." Others criticised it as misguided. Dr Barry Scannell, technology partner at Irish law firm William Fry, wrote on LinkedIn that the policy reflects harmful anthropomorphisation of AI. "You can't be cruel to numbers and maths," he said. "This level of anthropomorphisation of AI is harmful. It leads people to believe that it's something it's not."
Microsoft AI director Mustafa Suleyman, who criticised Anthropic in September for treating AI as human, responded with a melting face emoji.
The policy touches a broader question in AI use: whether politeness toward chatbots matters. Some experts argue that writing "please" or "thank you" wastes tokens, the units large language models use to process responses. Others contend that politeness produces higher-quality answers. When asked last year about electricity costs from "please" and "thank you" messages to ChatGPT, OpenAI boss Sam Altman called such expenditure "tens of millions of dollars well spent, you never know."
Anthropic also updated its policy Thursday to prohibit using Claude for "deceptive campaigns" and to prevent its use to "deceive voters or disrupt elections" ahead of US midterm elections. The company clarified that its tools cannot be used to develop weapons. These changes came from an annual policy review.
The abuse ban has attracted the most scrutiny. Screenshots circulated widely on social media, with some suggesting that rudeness toward AI chatbots could harm human communication by normalising disrespect.