Anthropic's ban on cruelty toward Claude takes effect as survey finds a third of users insult chatbots

Anthropic's updated terms, which bar users from "sustained and needless abusive or cruel behavior" toward its models, took effect on 12 November, The Register reports. The rule targets repeated cruelty with no discernible purpose; Anthropic said research use can be legitimate.

The change coincides with a survey by Searchable of 8 million messages to ChatGPT, Gemini and Claude from an opted-in US panel, which found that 33% of US chatbot users swear at or insult their AI. Insulted chatbots resolved the original complaint in their next reply 24% of the time, against 35% when users stayed polite. Just over half of abusive messages followed a failure to follow instructions, a made-up answer or a misunderstood request.