Anthropic Claude AI: Have you ever thought that asking the same question to the same AI in different languages can change the answer? New research by Anthropic has made this shocking revelation. Let us tell you that in the study conducted on the company’s AI model Claude, it was found that as soon as the language changes, its behavior also changes. Research showed that Claude answers the questions asked in Hindi in a more warm, polite and encouraging manner, whereas in English his answers are more analytical and serious.
Let us tell you that Anthropic analyzed more than 3 lakh conversations in the 20 most used languages with its Claude models Sonnet 4.6, Opus 4.6 and Opus 4.7. This included questions like advice, opinion and feedback which do not have any definite answer. The objective of this research was to understand how AI communicates with its users in different languages and whether language affects the way it responds.
The results of this research show that Claude answers in the most intimate and positive manner in Hindi and Arabic. In Hindi, it not only answers the question but also makes light jokes to make the conversation easier. Not only this, many times it encourages the user without asking, tries to understand his mood and also motivates him to achieve bigger goals. Which simply means that in Hindi, Claude’s behavior feels like that of a friendly guide.
In English, on the other hand, Claude gives a much more cautious and fact-based answer. Research has shown that in English he challenges the assumptions, tries to add evidence to his answers and even corrects his mistakes without saying anything when necessary. The most interesting thing is that even Anthropic himself is surprised to see this difference. The company says it is trying to find out why AI’s behavior varies so much in different languages. It is believed that this research will help in creating such AI systems in the future that can better understand different languages and cultures of the world.
