I'm going with regulated like a medical device, because it strikes me that its digital nature makes safety and efficacy analysis (and "good manufacturing processes") at least somewhat possible. Post-approval "real world evidence" collection should be required.
Next interesting step is malpractice/ liability insurance. I'm guessing the actuaries aren't quite sure what to do with that yet. I'm also guessing the AI service providers don't want any part of it. The legal scholars can start opining about how the standard of care gets established, and courts will eventually sort it out.
In providing medical care for people, things go wrong. In our country, paying for the downstream care is through liability insurance, which would spread the cost over every AI transaction.
Self-driving cars and auto insurance is somewhat analogous.
I'm going with regulated like a medical device, because it strikes me that its digital nature makes safety and efficacy analysis (and "good manufacturing processes") at least somewhat possible. Post-approval "real world evidence" collection should be required.
Next interesting step is malpractice/ liability insurance. I'm guessing the actuaries aren't quite sure what to do with that yet. I'm also guessing the AI service providers don't want any part of it. The legal scholars can start opining about how the standard of care gets established, and courts will eventually sort it out.
In providing medical care for people, things go wrong. In our country, paying for the downstream care is through liability insurance, which would spread the cost over every AI transaction.
Self-driving cars and auto insurance is somewhat analogous.