A third of US AI chatters are hot at bots, hurling insults at LLMs

A survey found that 33 per cent of US chatbot users swear at or insult AI assistants, prompting Anthropic to introduce a rule against sustained, needless cruelty towards its Claude model. The policy took effect on 12 November and reflects the company’s broader interest in how people treat AI systems, while distinguishing gratuitous abuse from interactions with a discernible purpose.

Searchable analysed eight million messages sent to ChatGPT, Gemini and Claude by an opted-in, anonymised US panel between September 2025 and September 2026. Chatbots resolved users’ initial complaints in 24 per cent of cases when insulted, compared with 35 per cent when users did not insult them; 52 per cent of abusive messages followed problems such as errors or misunderstood instructions, and 8 per cent came after refusals or safety warnings. Anthropic says its rule targets repeated cruelty without a clear purpose, while attempts to bypass safeguards remain prohibited under separate platform abuse rules.