The question is how AI uses the data. In theory, machine learning (actual current technology behind AI marketing name) is capable of generalization and coming up with new solutions not envisioned by the developers and their data. In practice, most existing applications collect so much data that when confronted with new data point, they simply need to find a closest data point in the data they collected. Not much generalization, just old-fashioned database lookup. See, for example, a self-driving car that gets further away in the rear mirror. It seems that they are still not able to collect enough data to resolve all possible situations.
Menu