There are two separate issues that require more work:
a) does AI potentially provide useful assistance in healthcare? While the suspicion is yes, it seems a little immature at present with some of the early claims being wound back due to having significant caveats and insufficient data (i.e. a lot of medical data is weighted towards white men). While I have no issues with small, incremental improvements, this doesn't justify wholesale access to medical records.
b) access to medical information should require informed consent with clear guidelines around exactly how can that information be used and, in the event it is misused or "lost", ensure adequate compensation is paid to those who are affected. For instance if medical data with PII is provided to insurance companies.
While point (b) makes point (a) harder, I'm not convinced at this point that the benefits justify any potential risks of mis-use. What is the point in having a medical issue that is treatable due to AI when you can't get insurance to cover the treatment?