I also worry about anything that's attempting to track the white lines/dashes at side of road, and uses that to guide the car. Tesla had a fatal when their very sophisticated camera system followed a triangular area with shaded white lines that was getting wider. It followed the left edge of this area, which was the slip-lane exit off dual carriageway wih bridge coming up ahead. The body of the car drove onto the ever-widening hatched area, then ploughed into an energy-absorbing crash-barrier protecting the concrete pillar of the bridge. Unfortunately, this crash barrier had already been flattened by pevious impacts, so no energy got absorbed in this case. Tragic.
I am not yet convinced that VW's camera system is necessarily more sophisticated/well developed than Tesla's was back then, and I'm certainly not going to trust my life or hand over any driving esponsibility to that particular feature. Happy to have stuff that adds to my abilities, radar ACC etc, those don't really take away the need to concentrate, and I can't see them doing anyhing too dangerous (unless it brakes unexpectedly, but tha probably means I'm doing somehing wrong to begin with).
Image processing & line following is not a trivial task. Add in broken/warn lines, etc, just makes it exponentially harder.
Some years back, a US project tried ot get an autonomous vehicle to follow a path between 2 well defined lines. Front-facing camera took a picture of what was ahead, looked for two sharp edges, one on left, one on right, converging slightly, and steered the vehicle between the 2 lines. Easy task to follow a perfectly marked one-way street with no obstacles whatsoever. Even I could code that one up!
They tested this out in the open air somewhere, roughish terrain (I think this was a military off-road thing). Camera saw a tall tree with no branches to speak of in frame, edge-detected it, got 2 nice converging edges, and drove straight into the tree!
It's the same issue as Tesla have. How do you distinguish what's a proper line, from what's been painted in error, or painted correctly but with a different meaning than expected. You need a very, very sophisticated model in the car's brain to handle this stuff.