it will do whatever its programmers tell it to do, based on the data and models it is supplied with.

Whatever it returns will be based on whatever source data its been fed, and the accuracy thereof. Which leaves it open to bias of any sort, of spreading misinformation, etc, if that has been included in its source feed, whether deliberately or unwittingly.

However, if there is an effort to sanitise all of those data feeds, there is also a risk of introducing bias from that sanitisation process based on the chosen sources