This seems to be an argument in favor of Tesla's full-stack approach of pairing high-capability infotainment with advanced driver assistance features. Tesla's data on collisions suggests the driver assistance works to reduce incidents.
Using Apple CarPlay or Android Auto to add higher-capability infotainment to an otherwise unassisted vehicle does seem obviously risky and the data from the linked study appear to confirm that. My own anecdotal experience is that I am very distracted by Apple CarPlay in rental cars (my own vehicle is too old to be compatible). This is especially because the user interface is slow, with too much emphasis on animation rather than immediacy, and with too many behaviors that require waiting, such as resetting the map back to full-size.
Using Apple CarPlay or Android Auto to add higher-capability infotainment to an otherwise unassisted vehicle does seem obviously risky and the data from the linked study appear to confirm that. My own anecdotal experience is that I am very distracted by Apple CarPlay in rental cars (my own vehicle is too old to be compatible). This is especially because the user interface is slow, with too much emphasis on animation rather than immediacy, and with too many behaviors that require waiting, such as resetting the map back to full-size.