And presumably, AI adds other features.

Yes, it adds a layer of additional complexity to an already difficult question to answer.

I think when you cut through it, what it comes down to is a medical device is really defined by what it is meant to do.

Put simply, if it is intended to be a medical device, it should be regulated as well.

That in the current system is already quite tricky for the reasons you've set out.

But with AI-enabled products, it can really be a difficult task to properly characterize what a product's intended purpose is, particularly when its capability, design, and risk level keeps evolving over its lifetime on a constant basis.

So what this report has done is say that the rules governing what is and what is not a medical device need modernizing to make it clear, for example, when software that's intended for administrative purposes but in a healthcare setting or general well-being apps or some other low-risk decision aids or tools, to make it clear when they are not medical devices.

How do you regulate a product that keeps changing?