In terms of electronics it is fairly trivial and it can be inspected by eye (or microscope) if the manufacturer decides to not encapsulate everything on a chip (which presumably would be the point of such a feature).

Just have the only positive voltage rail going to the camera be the same one that is directly powering the LED. The firmware will be turning this rail on and off, hence turning the camera and the LED on and off simultaneously.

Idle though on possible attack vectors:

Convince the firmware to use a lower voltage, one that doesn't hit the breakover voltage on the LED but still powers the camera.

Strobe the line, get snapshots without the LED doing more than very faintly glowing.

I have a very hard time believing that the camera itself will not brownout in such a case.

LED activation voltage is less than what cameras typically require.

Neat ideas. It never even occurred to me that the power rail would be voltage-programmable, but of course it probably would.

Your second idea is even more scary and realistic.

Even if it was, most of the time when people are taking surveillance webcam photos (a hacker or a person trying to recover their stolen laptop) you can take the picture so quickly that they probably won't see the light.

Have you a source for that? My experience of laptop webcams is that they need at least a second or two after power on to initialize the camera. That'd be pretty noticeable if you're using the machine, at least.

I do! DEF CON 23 - Ken Westin - Confessions of a Professional Cyber Stalker

https://www.youtube.com/watch?v=Tyel1gfSjl4

Specifically around the 18:00 mark. Also he does provide source examples that you can ruin your computer with if you want.

He gets into a lot of it, but shows quite a few examples including what I'm talking about. The people that he catches aren't necessarily savvy, but he does talk about taking pictures fast enough for the light not to be visible.

Thanks!