So long as they're totally de-identified (not linkable to people), this is a good thing. Image a GPT that can diagnose you 10x as accurately as a human and 1000x faster. This can revolutionize medicine.
Generally, data science is evil and so is Google, but this needs to be done.
That is a greater good argument, but who really benefits here? In a somewhat best case scenario, some big companies and their equity owners gain all the data and perhaps develop proprietary solutions only they can prescribe. We then sacrifice our collective health privacy for some people who might benefit if they pay an uncompetitive price for the treatment. In a worse case situation, no cures or treatments are ever found, but these companies benefit from the surveillance and knowing what existing drugs and treatments they should invest money in to profit from, allowing them to game the market.
If this data was held in the commons and the benefits such as cures or treatments were not patented, we might have some incentive to provide it. In this case, everything is privatized for the benefit of a few.
I believe there's potential for great benefit for everyone.
I also believe that if my records are used to innovate I should get a cut and get an opportunity to opt in or out. I should have agency. I don't like that I don't.
I still believe that this is, on balance, a good thing.
Bold of you to assume "we" have electronic patient records here in Germany, when they just introduced the e-doctors certificate and e-prescriptions a few months ago and it's already not working as intended.
IBM did it first. It wasn't a secret. There was a Watson Health group dedicated to training ML models on medical records from large insurers and hospital networks. Among other things, the game plan was to have the system provide oversight for the notes of physicians and other medical practitioners - to spot poor quality/repetitive notes and alert the practitioner and/or their boss to the risk of malpractice/inability to bill for the encounter.