How could an AGI have legal protections in an our current society? [closed]

Question

First: assume society is as it is today, British common law

Many “stories” try to package AI as citizens (DATA of Star Trek, Sonny of I Robot). They manipulate society to make AI fit with some certain directives.

This is not that question.

1. This is NOT asking how to make AGI into citizens. A simple law does that, it’s not the question I asked.

Laws of society can only grant rights if there is some means to remedy the violation of the right. For example, a law that protects a human’s right to breathe water can’t be tried in court even if some human finds out they can’t breathe water. The law simply can not possibly “fix” the thing that’s been “taken” from you - an impossible ability to breathe water. That is a ridiculous example, I know, but ridiculous laws do happen. The point is, simply saying you have a “right” to this or that doesn’t create the legal right. The legal right to have anything only exists when some way exists to repay you when it is lost (a remedy). The law originated in Roman law as ubi jus ibi remedium, “where there is a right, there is a remedy”, and remains in effect today. A real-world example: we have a right to have our credit information accurately reported. This law is called the Fair Credit Reporting Act (FCRA). But even if the credit company gives false information about your credit, the court can’t protect you. Spokeo, Inc. v. Robins (2016) ruled that even though misleading figures were published by the credit card company, the plaintiff could not show that they actually lost anything from inaccurate or incomplete information in their credit reports. There is no right to relief until you have actually “lost” something.

2. I am not asking how to change any laws to accommodate current AI.

Because AI is not AGI. This is about designing a fictional Artificial General Intelligence (AGI), and answering what “ingredient” - for lack of a better word - would afford it rights under existing common law: Let’s assume we invent the asked fictional AGI, and we put them into a mechanical body. Now let’s assume a group lobbies and passes a law that makes it illegal to willfully dismember and destroy the AGI, and treat such an act as a “wrongful death.” They classify this as a crime just like it would be for a human (because again, no laws are changed)

When an AGI is destroyed, someone claiming a relationship to it wants the system to prosecute for wrongful death. Well, there clearly exists a “wrongful death” law in the books because it was passed. But the judge’s first job is to decide if some remedy exists to what was “lost.” This means the judge has to be convinced that at first, the AGI even had a right to life before asking if a life has been lost.

So the question: If an AGI is a computer program, and computer programs can be and are backed up and saved regularly, so effectively they can only be lost by deliberate manipulation of the server; does a computer program have a “right to life?”

What about our world (specifically, about the qualities of an AGI in this world) needs to be changed

to give algorithms and programs a right to life, that could be recognized and fairly remedied in a human justice system? (Answers do not need to fix the problem, the question only asks what needs fixing)

Emphasis again, I don’t believe our real world could possibly argue for the right of an artificial construct to exist, as they currently exist. The question tries to pin down what would need to be different so that it could (by changing the AGI, or maybe the environment - anything except the basic principles of jurisprudence)

I can’t state this enough. This question is NOT asking about current AI or anything existing today.

It’s a fictional AGI that can fit into our society; why does it fit? (e.g., because it can reciprocate?)

Answer 1

Interesting question - but I think the thinking here is misguided.

Assume for the moment our AGI is a full actor within our society. That is, they are free to work as they see fit - a completely autonomous entity.

Let's also assume that there's a perfect backup of them at all times.

The question isn't whether the AGI is alive, it's whether or not they have a right to conduct their affairs free from interference. Think Freedom of Movement, Freedom of speech and bodily autonomy. The right to life is merely an extension of the right to Bodily autonomy - no one else gets to decide what is done to your body.

If someone interferes with our AGI, even if they are not legally alive they have had these rights interfered with. A Malicious act done against them stops them from what they otherwise would have been doing and as such, a wrong has definitely been committed.

Since you mentioned 'wrongful death' - you are talking Civil and not Criminal law - in which case you may not even need 'wrongful death' - Tortious interference (I think, I'm not a lawyer) should cover pretty much everything you need.

But in your other section, it sounds like you want it to be Criminal law.

That all said though - here is where a Judge would reason:

"Restoring from a Backup is not a guaranteed process, there is a risk every time Data is transferred and restored. The Legal archives are complete with cases of Manslaughter where something that would ordinarily be survivable, lead directly to a persons death and the individual responsible found themselves guilty of that Crime.

As such, it is no defence to say that no injury has occurred because a restoration from a backup can or was done, the injury was done when an self-actualized individual was interferred with against their will - and as such - a crime has been committed"

Answer 2

Nothing under current law allows AGI rights - and can't.

Under current law, Humans have rights. Animals also have some rights, where humans stand in as the litigators - either as the owner of the litigator, the state, or as friends of the animals.

AGI does not qualify under any law for anything other than property. In fact, it does not feel pain, shame, or suffer injury in any way resembling humans or animals. As a result, it is nigh impossible to apply a vast majority of laws to them directly.

There are laws in effect that pretty much ban giving AGI rights, especially property and copyright laws. In an evil twist, due to copyright, the AI does not have any right to replicate its own code, and thus it can't qualify as life /in any way... what is an AGI anyway?

• It does not have the capacity to reproduce.
• It does not die on its own.
• It does eat electricity to live.
• It exists inside a computer.

That only allows concluding that it is software, not life in the traditional sense. Software is property under current law. Someone owns the source code as intellectual property, everyone else owns licenses to use the program. This directly contradicts with the AGI having any rights: If it had rights, a legal system that does so would need to allow slavery. currently, slavery is outlawed worldwide.

Due to property law, the AGI doesn't even own (or pay for) the servers it runs on. If it had rights, it also would instantly be trespassing on the servers, and be evictable. It couldn't go anywhere, so eviction would mean this particular copy of the AGI's code would be destined to die. Which would cause problems with evicting the AGI. Or shutting down power to the AGI's servers - which would follow the moment that an AGI somehow legally would own a server but could not pay for the electricity bill.

Oh, and a silly side note: AGI in themselves were trained by possibly violating copyrights - some courts are currently testing if training an AGI with the works of an artist or writer constitutes copyright infringement, which can mean that the very existence of the AGI is a violation of law in itself! This means, under current law, the AGI could need to be reset and untrained due to being trained on material that is under copyright.

For more about how the real world treats AI, look at the artificial-Intelligence tag on LAW.SE

Conclusion: Outlook too hazy

You need to invent a fully new branch of laws that only deals with AI rights and that does not violate any of the concurrently existing laws. That will be nigh impossible from today's point of view.

So, as we would need to reinvent all the laws in the first place, we can't properly speculate how those laws need to look in a sensible manner. You will need to look into literally all branches of law (save traffic law), especially property, tenancy, and intellectual property law, as well as the various torts of criminal law and free speech law.

You have put up a monstrous task trying to give AGI (not-really-AI) any rights - because with a real, true AI, which might be bound to very specific hardware and can't be copied truly, you at least could say, that that hardware is its body, that it owns that piece of hardware - and still get into the mess of turning off power to it.

Until you have figured out how Intellectual Property works in your world, and then figured out how access to or possession of computers is handled, you should not look into the rights situation of AGI at all.

In other Media

If you want inspiration for proper AI rights, look at the following case studies:

• TOS: The Ultimate Computer - A tactical computer on rampage
• TNG: The Measure of a Man - A True AI on trial if it has rights
• The Multivac Stories - A cycle of Asimov stories in which a computer with (nigh) true AI runs the governments. Some of them delve a tiny bit into the rights of the AI angle. Usually, they treat the computer and its AI as the same entity and not a true person with rights.
• The Robot Series - Asimov short stories and novellas. Among the topics: Can a Robot be a killer? Are the Three Laws Safe? Robot rights are touched upon multiple times and can be read as AI rights discussions.

Answer 3

There certainly could be legal protections developed, modeled on existing protections for humans, but there would be important distinctions which would need to be decided.

Question 1: What constitutes an AGI's 'self'? If you make a perfect backup copy of an AGI, is it considered the same entity? If the running instance of the AGI continues to learn and grow while the backup remains static, is there any point at which the differences between it and the running version are so great that they become distinct entities?

Similarly: if you make a perfect duplicate of a human, and throw them in cryogenic stasis while the original continues to live their own life, is the duplicate considered a separate person? I think most people, myself included, would consider the duplicate human to be a distinct individual as soon as they're created.

Potential Answer: Each legal instance of an AGI is considered a unique individual. Legally, they can create static backups, but those backups are not considered "alive" until they actually begin running and therefore deviate from the original. Any running duplicates an AGI creates without authorization are considered illegal, and have no protection against deletion - the purpose being to prevent an AGI from behaving as a virus and creating numerous duplicates which no one is allowed to delete. It also allows an AGI to create a temporary duplicate, and then delete them themselves without committing "murder".

Question 2: What constitutes "death" for an AGI? If you delete it, it's obviously destroyed, but what about deleting only part of it, such as its memory of a certain period of time? Is it still technically the same entity if part of its data has been manipulated? If so, then how much alteration would be possible before it is considered a different entity, necessarily indicating the death of the previous iteration?

Potential Answer: Any modification to an AGI, either a running instance or static backup, beyond the scope of its allowed inputs, is considered identical to maiming/mutilating it. And, if a prosecutor can convince a jury, murder of the original entity. The "scope of its allowed inputs" clause basically means that it's ok to tell ChatGPT via its text interface that up is down, and trying to convince it that this is true; it's not ok to edit its code or running memory to make it believe it hat.

Answer 4

An AGI is a computer program. Yes, computer programs and their data can be backed up, so if an AGI was housed in a mobile computer, i.e. a robot chassis, and that robot was destroyed, the AGI and its data could be restored to a new chassis as of the last backup.

If a robot housing an AGI (or even if it does not house an AGI) is destroyed as the result of unlawful action (or an insurable event), then there is a financial cost involved in replacing the robot chassis and reinstalling the AGI and its data. Legally, the entity responsible for that unlawful destruction (or responsible for providing insurance) must pay that cost and any other penalties a court might impose.

The question of legal rights for AGIs is irrelevant to this so far.

Now we come to AGI rights. If AGI has no rights, then each instance of an AGI is owned by some entity with legal rights. If AGI has legal rights, then the AGI may own its robots.

However, the nature of AGI as a computer program means that unlike humans, its existence is not necessarily tied to its physical form. Provided that it was backed up, an AGI would be able to replace a completely destroyed robot with the loss of only time and perhaps some data.

So, assuming that AGI has 'human' rights, destroying a robot would not necessarily mean destroying the controlling AGI. However, since an AGI is effectively data, that means that the possibility exists that the AGI itself could be destroyed, if its robot chassis and its backups were all destroyed.

In the case that an AGI (not just its robot chassis) with human rights was unlawfully destroyed, it would logically be considered to be the same as the unlawful killing of a human.

That means that the perpetrator(s) of the unlawful destruction of the AGI would be prosecuted for that destruction, theoretically with penalties equivalent to that of killing of a human being applied if found guilty. Given that AGIs would be so easily backed up, actual destruction of an AGI and all its backups would require deliberate actions be taken by the perpetrators, and/or wilful negligence by the AGI. Those factors would be considered by a court.

As for the other consequences of the actual destruction of an AGI with human rights, they would be similar to those when a human dies: their property would be assigned to their designated heirs, or in the absence of heirs, would become the property of the state. Any damages or compensation payable by the perpetrators of the unlawful destruction would also go to the AGI's heirs, or the state.

In the event of the unlawful killing of a human it is a general legal principle that the state punishes the perpetrators of such actions, and that those with emotional or financial links to the deceased may bring civil suits against the perpetrators if the state does not claim prior interest. If an AGI had human rights, the same legal principles should apply.

Therefore, a world with human rights for AGI would be little different to one without. The only questions might become, 'Did you know that the entity you destroyed was a robot and not a human being?' since the penalties that would apply would depend upon the perpetrators' knowledge of the consequences of their actions.

Now, to extend these principles, we need to consider the consequences of unlawful damage to a robot or to an AGI and its data and backups.

Again, the tried and tested principles of the value of the loss and its consequences would apply. Destroy a robot's leg? How much would it cost to make good, and what is the cost of the consequences that followed? Modify an AGI's algorithm or data unlawfully? The same principles would apply. That the consequences affected or even may have affected others should also be considered.

Answer 5

It depends

A very interesting question, the answer to which, however, strongly depends on the legal system. The question itself is obviously aimed at the common law system prevailing in English-speaking countries.

Under the variants of the civil law system prevalent in most non-English speaking countries, the specific question would not arise (and such a concept as "wrongful death" is not transferable to these legal systems either). If there is a law declaring something to be life, it is legally to be treated as life (and can be killed at all), judges and courts would then be legally bound by that.

That is why the development of AI rights in different countries would be highly different and at different speeds, would be my guess.

Development of animal rights as a historical comparison?

An apt historical comparison would perhaps be the development of animal protection/animal rights. Until a few decades ago, animals - regardless of the legal system - were exclusively a subject of property law. In many countries, however, the law has evolved; in Germany, for example, animals have a kind of quasi-subjective legal status. They are still mostly property (pets, livestock, etc.), but their protection has had explicit constitutional status since 2002. The Animal Protection Act, amended accordingly, defines the principle "No one may inflict pain, suffering or harm on an animal without reasonable cause". In practice, this means (in criminal law) that in the case of unlawful killing of an animal, not only the violation of property is punished, but also the killing itself - which regularly leads to a higher sentence. It is even the case that I am not allowed to kill an animal in my property "without reasonable cause" either, but can be punished for it. In civil law it leads to the fact that in case of unlawful killing of an animal a higher compensation can be claimed than in case of destruction of non-living property.

My prediction: if something like AGI were ever to emerge (which is not the debate here), AI rights would also emerge only gradually.

Completely different from the development of animal rights, however, would be that an AGI would probably be quite capable of representing its rights itself. In the mentioned scenario, the backup could well take action itself and take the destruction of the original to court.

When will the backup become another person?

Whether the destruction of the original AGI is to be classified as killing depends on the frequency of the backup (and the subjective time perception of the AGI!). If the backup is stored redundantly, e.g. as with hard disks in a RAID array, then probably not. However, if the backups should only take place in larger time intervals, the question quickly becomes philosophical: Since each experience has influence on the personality, the question would have to be clarified, from which time lapse two different persons must be assumed. Given that an AGI would probably experience time subjectively much faster than a human being, I tend to assume that with non-redundant backups it must be considered very quickly that we are dealing with a different person than the original.

Legal guardian for AGI?

A word about "ubi jus ibi remedium": I do not think that this principle can be universalized anymore. In Roman law, children were property, which they no longer are in (almost) all legal systems. Undoubtedly, children have rights, but they cannot yet represent them themselves, and as infants they are not even aware of them. This is where legal representation (in the case of children, by their parents or legal guardians) comes into play. So, in the context of AGI, the question would be: Are AGI minors? And who is then the legal representative?

It would be a mess

So, to answer the final question: Such a world would differ from ours primarily in that everything would be an even bigger legal mess than it already is. Good times for lawyers, most likely.

Answer 6

Laws and rights are artifacts made by humans for humans. Those who have tried to use laws to enforce nature have failed.

What needs to change is that the law has to recognize that AGI is a bearer of rights and maybe duties, and under that legislative umbrella everything can follow.

It has already happened in the past:

• when slaves where recognized has having a soul, they could no longer be seen as objects, but had to be seen and treated, if possible, as humans (which didn't happen immediately, but at least the principle of law was stated).
• when women, analphabets, poor people, etc. were not given the right to vote, the law was changed to make it possible.

I think in our real world there are movements trying to have rights recognized also to animals.

What you need to have is a movement of opinion which sees giving rights to AGI as positive, and that movement must have enough support to be able to influence the legislative power.