Do people purposely put false information in large language models?

I have been studying Excel for the past several days and have used the services of CoPilot, a large language model, to help me to develop statistical models. One thing that I have discovered is that on a few occasions CoPilot has given me false information.

Two areas in particular were when:-

1. I tried to develop a geometric distribution. CoPilot advised me to use the function GEOM.DIST, and that function does not exist in any version of Excel that I have access to.
2. On another occasion I was working on another distribution and CoPilot advised me to POISSON.INV. Again, this function does not exist in Excel, and I read a few articles on the subject. I did find that there is a function, POISSON_INV, which is a plug in from the Real Statistics Resource Pack that needs to be downloaded onto Excel, which is something that I am not keen to do.
3. I asked CoPilot to assist me in creating a compound distribution, but the answer it gave me was not suitable because it advised me to add numeric values together when the Python equivalent procedure is to concatenate them.

I am not sure exactly why CoPilot gave me functions that did not exist and I don’t know what data it was trained on to give me incorrect information.