You didn't watch the movie good. The whole point was that they weren't 100 percent accurate and things were not only wrong but could also be manipulated. So also made up bullshit.
They could accurately predict the future. The issue was that they didn't always agree. When they didn't agree the dissenting opinion was filed in a minority report and they'd go with the majority opinion. That was the problem, that they were arresting people when there was more than one possible outcome, lying to the public, and covering it up.
I, for one, am glad that AI is no where near able to implement this ability. That said, I'd be worried if they said they implemented it and used it as a way to target political enemies.
maybe the person in charge of the program (let's call it the Pre-Crime program), is suddenly tagged as the next person to do a crime!! Suddenly he has to get out of the PreCrime facility in Buenos Aires as quickly and quietly as possible. He quickly finds the programmer that designed the PreCrime AI system and learns that the three AI units don't always agree, and the dissenting AI creates a minority report that is hidden from the records by default.