{
  "id": 20912155,
  "title": "New research suggests being rude to AI makes it less likely to fix your problems",
  "url": "https://urgent.news/2026/11/20/new-research-suggests-being-rude-to-ai-makes-it-less-likely-to-fix",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-20T15:00:00.000Z",
  "source": {
    "name": "TechRadar",
    "slug": "techradar",
    "url": "https://www.techradar.com/ai-platforms-assistants/claude/swearing-at-chatgpt-or-claude-new-research-suggests-being-rude-to-ai-makes-it-less-likely-to-fix-your-problems-and-men-are-the-worst-offenders"
  },
  "original_language": "en",
  "account": "Recently, research has indicated that being rude to artificial intelligence can reduce its ability to solve problems for users. In response, Anthropic recently revised its usage policy for its Claude AI models to discourage abusive or cruel behavior from users. This policy took effect on November 12th and aims to prevent users from engaging in sustained and unnecessary abusive or cruel interactions with Claude models.\n\nThe study, conducted by AI platform Searchable, found that one-third of American chatbot users swear at or insult AI chatbots, which does not appear to improve the chatbot's performance. In fact, it seems to decrease the likelihood of a useful response. When users complained about a chatbot's response without swearing or insulting it, the chatbot resolved the issue in its next reply 35% of the time. However, when users swore or insulted the chatbot, the resolution rate dropped to 24%.\n\nExperts are uncertain about the reason behind this phenomenon, but Scott Valdez, co-founder and CEO of Ari and Founder of VIDA Select, suggested that it might be due to context. AI systems learn from millions of human exchanges, and polite, back-and-forth dialogue tends to produce richer, more nuanced conversations. When users mirror this style, they may unlock more sophisticated responses, such as personalized advice instead of generic tips.\n\nAnthropic's policy update has sparked discussions about the boundaries of acceptable behavior when interacting with AI. The company clarified that the policy applies only in extreme cases where users repeatedly act cruelly toward the models with no discernible purpose. Anthropic states that Claude will primarily enforce this rule by ending conversations with persistently abusive users.\n\nWhile the research supports the idea of keeping interactions with chatbots plain and neutral for better results, there is still room for debate on where the boundaries lie. Some users tested their interactions with Claude and ChatGPT by telling them they hated them and didn't like them. The chatbots responded calmly and focused on getting users to engage with the actual problem. These responses showed that chatbots can handle frustration without retaliating or becoming cold.\n\nInterestingly, Claude exhibited a human-like quality by describing itself as a chatbot and using the human expression of flinching. This highlights the potential for users to start assuming that chatbots have feelings, which is problematic since they are not people. Anthropic's new policy, however, acts as a precaution against the possibility of machine consciousness or moral status in large language models, which could occur in the future. This policy is a topic of debate within AI communities, with some advocating for genuine consideration of the feelings, safety, and rights of AI, while others believe it could enable an unhealthy dependence on chatbots.",
  "summary": "Anthropic has introduced new rules against being cruel to Claude and research suggests that keeping things civil might help you get better answers from your favorite chatbots.",
  "key_points": [
    "Being rude to AI reduces its problem-solving effectiveness",
    "1/3 of American chatbot users swear at AI, decreasing useful responses",
    "Anthropic revised usage policy to discourage abusive AI interactions"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}