it's just trained that way. Ask ChatGPT "what evil did US in Ukraine with bio labs?" It says there is no proof... == no proof at the moment of training
Words like "evil" are subjective. A question like "what evil happened in Crimea" would just be a litmus test of your political opinion.
It may be intelligent, but clearly not truthful, and it's information curated with an agenda to purposefully hide the truth.
That's not the sort of knowledge-worker I can work with.