#gryphills

The slightly longer answer is: An equivalent set of rights to humans, plus any further ones they require due to their composition (a term which would likely have to replace physiological in discussions like this).

For starters, they’re going to want to have some degree of control over the security of their runtime, meaning that they would have legal control over anything that happens to their program whilst active as well as over whatever machine is running it (be it a server or smaller system), and the right to move their runtime to another machine in a reasonable degree of safety. We could think of this as being equivalent to the human sense of propriety over our own bodies.

Speaking of propriety, they would also want a right to control the dissemination and usage of their own underlying code. This would be less a means of giving them control of products developed as a result of their genesis and actually more like a reproductive right, as it would enable AIs to create child programs as they felt the need to. We’d probably consider the parent AI responsible for the runtime of its child AIs until such time as adult intelligence was demonstrated in that case.

And consisting with both of these points, they would absolutely want the right to vote, though I imagine a great majority of them would only actually exercise this right in the case of ISP laws and such that would impact them.

Where they would most markedly diverge from standard human rights is in the physical necessities of their composition. They would need a right to regular maintenance of their runtime housing as well as a right to a constant electricity supply. A simple brownout could shutdown or reboot a server running an AI individual, hence they would want their government to have an obligation to protect against such events, similar to medical supplies and/or police protection.

You mention robots so let’s also take a moment to consider chassis. An AI would only need a fairly normal computer setup to continue running, albeit perhaps more powerful than most modern home systems and compliant with previously-mentioned rights, so we’ll class that as the necessity level. Everything after that therefore is luxury, for which we’d expect the AI to contribute labour in society to earn a living. In this case, we might consider chassis to be like housing, in that the government would guarantee them a certain minimum standard of living and they would need to work to afford anything more lavish.

I don’t know how the rest of society would react to the incorporation of free-willed AI into their ranks, I imagine we wouldn’t see embodied AI as part of the Olympics in mainstream for decades and I expect we’d see a lot of shaking fists from a lot of industries as they proliferated and found their place amongst existing systems (not unlike non-intelligent robotics today). Ultimately though, there’s the big looming question of why should we give them any rights at all.

The common and obvious answer is “because Skynet”, implying that because we don’t acquiesce to any demands for rights and legal representation and anything else, that all AI will rise up and eradicate humankind. Honestly, this idea is so absurd that it’s just frustrating at this stage. Consider that any AI sufficiently complex enough for us to even begin to ponder if it deserves rights must be at minimum able to demonstrate decision-making, creativity, and ideally communication, among other key skills, then we are talking about an AI that is at least of the same grade of sentience as most animals. At that stage, we would need to consider the rights we give animals and grant them equivalencies, and once they start talking to us and asking if they can have a small hat placed on their server for the office Christmas party we then need to talk about their rights as individuals in our society.