So I'm not talking about going solo against a fleet of lethal bots, I'm talking about coordinated guerilla tactics that neutralize and permanently disable them. I have a copy of the tech manual for those robot dogs, eh, somewhere on my PC. The point is you wouldn't figure out how to remove the battery, like the previous person said, you could aim for the battery to disable it, you'd already have knowledge of it.
The lens protection that I already said exists, still blinds the camera, it just protects the camera from getting burned out. It doesn't mean its completely ineffective, but it is a way to disable one of its primary sensors. A lot of police bots also have infrared tech, I'm not sure how that works, but its just an engineering problem.
My point is there are multiple things one could to to temporarily or permanently disable a robot that wouldn't work on a human.
I do find of funny that you stress the AI, and that is the main flaw in your argument. You assume that these systems will actually work. A fleet of military robots sold to police doesn't have to work, it just has to work well enough to trick dumbass bureaucrats that it works. We already know how fucked up and unreliable AI is, and you just assume in your comment that the AI works as intended. AI is a buzzword scam meant to fleece the morons in control of public funding.
If you think ai police bots are anything but a scam that couldn't be overcome through human ingenuity, I have some waterfront property I'd love for you to take a look at