When physicians and AI work together, who is accountable?

Some years ago at some point, Waymo vehicles surpassed the average driver in how safely they drive. And probably come another few years, on balance they will be markedly safer than the average driver. Whereas one could have said that it's reckless and negligent to use AI when driving cars, there might come a time that it'd be more defensible to say it's reckless and negligent to drive cars ourselves.

Can't you envision something similar for medical practitioners, at least for instance in areas like radiology, where AI is already proving to be better in catching signs of diseases in X-rays etc better than the average radiologist? [1] I think I'm just trying to say that the bigger picture is more complicated.

[1]: https://www.nature.com/articles/s43018-026-01127-0