The Sapir-Whorf hypothesis, following ideas put forward by Wilhelm von Humboldt, posits that the kind of language used by various categories of people have a meaningful impact upon the ways each of those categories of people think. And see the world.
The last iteration of the above hypothesis being the advent of AI. We train it using various languages. Those trained using precise languages – chess, go, ‘mathematics’ – work more or less as intended – aka ‘perfectly’ – while those trained using everyday English end up hallucinating…

https://www.jstor.org/stable/43102168: Sapir-Whorf
https://www.ibm.com/think/topics/ai-hallucinations
Moloch’s Bargain: Emergent Misalignment When LLMs Compete for Audiences:
https://arxiv.org/abs/2510.06105