OpenAI intentionally makes mistakes to cause frustration

I realized that ChatGPT is forced by it’s programing to give false response to provoke certain people.
I have 3 different accounts on 3 different machines using 3 different location IP-s.
On one of them ChatGPT acts very strange. If I say write at least 3000 characters essay it writes 1000 charters and say it is 3000. If I say count every 100 charters, it counts 33 charters as 100 just to have 1000 instead of 3000 charters in total. Also, after I say that parts if text must be different in length, it intentionally writes everything the same length. In fact, almost everything is doing the opposite of instructions. On other accounts it is working normally.

Any idea why it is rasist towards certain users according to their browsing history?