I would have thought this would've been asked already, but I searched this site and the Internet, and I couldn't find anything. Apologies if this is a duplicate.

As computers become increasingly autonomous, there have already been a handful of incidents where computers have done things that would be criminal for a person to do. For example, a self-driving car killed a pedestrian, and a Twitter bot made a death threat. Moreover, both of these computers were essentially operating under their own decision-making processes and were not directly controlled by any human.

Could it be possible for prosecutors to file criminal charges against a computer? Is there some standard of sentience that the computer would have to achieve in order to indict it? (United States)

  • 5
    You are talking about science fiction, not law. Oct 20, 2019 at 0:11
  • 2
    I think if someone makes a program that “might” break the law and the programmers are aware of this possibility, they as humans are 100% liable - to program, release or run it knowingly. Oct 20, 2019 at 3:43
  • 3
    This is no more science fiction than most of the law surrounding technology. What prevents a particular physical computer or identifiable abstract programme from being indicted, when physical rivers and abstract companies are granted personhood and can (to varying frequency) actually prosecuted per se? @GeorgeWhite
    – user4657
    Oct 20, 2019 at 4:27
  • 4
    @GeorgeWhite a self-driving car killed a pedestrian. That is fact, not fiction, and surely the law has something to say about liability in connection with that death. Anon, I am unfamiliar with the death threat you speak of. Could you add some links to the question for the benefit of readers who might not be familiar with either of these incidents?
    – phoog
    Oct 20, 2019 at 6:34
  • 1
    A computer driven car will not be going to jail any time soon no matter how many people it runs over. Oct 20, 2019 at 19:31

3 Answers 3


First, I will make a distinction of terminology.

A computer is hardware, like, in essence a calculator. Hardware doesn't make choices or decisions. Its response to any input or occurrence is dictated by the code it is running (barring misfunction). It's behavior to a given input is generally static or in set in a pre-specified pattern (e.g. press a button, show 1+the last number shown). However, it often responds to an input by giving inputs to a program, and then following its instructions (like a car, responding to a driver's input of pressing the accelerator peddle by speeding up).

A program is software. Software and its behavior can be edited, and can even alter its own behavior (see the topics of "machine learning" and "artificial intelligence", and note that in real life they are actually quite different from how they are commonly portrayed in fiction, much like law).

With regards to a computer, technically it could be targeted directly by legal action today under the rule "civil asset forfeiture". However, in general a computer is a merely a location where programs are stored and run, and so are unlikely to be indictable in the traditional sense, in the same way that one would usually not be able to indict a field or building.

Potentially, a program could be indicted, if a law was passed indicating that programs or classes of programs are legally considered persons. Note that either this only apply to programs that can pass some series of thresholds or would have massive additional effects because programs would then be full persons, under the law; so you wouldn't be able to buy, sell or license property (due to slavery laws), nor shut down computers (because doing so would "kill" the programs that are stored there (in the same way one can't burn down a building where people are living), nor alter programs without their consent (analogous to medical procedures).

  • 2
    If a computer was a "person" like a corporation was a person, it could be bought/sold and essentially killed. Oct 20, 2019 at 20:37
  • But if a computer was a "person" like a corporation then it couldn't be charged with murder or put into jail.
    – gnasher729
    May 3, 2021 at 19:45
  • @gnasher729: It can't be put in jail, no. It can be fined, have its ability to operate in specific jurisdictions or ways curtailed, or be shut down (i.e. put to death). Not only can companies be charged with crimes, they can also be convicted of them, including homicidal crimes. Or at least manslaughter, if not murder, but BP apparently has plead guilty to it(en.wikipedia.org/wiki/…), apparently related to the Gulf of Mexico oil spill.
    – sharur
    May 3, 2021 at 20:00

Could it be possible for prosecutors to file criminal charges against a computer?


Is there some standard of sentience that the computer would have to achieve in order to indict it?


Of course, some jurisdiction (e.g. a U.S. state) could always change the law.

Criminal charges resulting in the imposition of fines and collateral consequences of a conviction can be imposed against corporations. But they cannot be brought against things or animals, no matter how intelligent. A computer could be the subject of a civil forfeiture, but that isn't analogous to a conviction of a crime.

It is possible to bring a John Doe criminal indictment, however, only to drop charges when, contrary to the implied assumption of the indictment, the crime wasn't committed by a person.

One of the standard trick questions in criminal law is to present a thing or an animal in a context where you want to think it was murdered (which you can't in the case of a thing or animal victim), or was a perpetrator of a crime (which it can't be).


Consider Kipling's "Hymn of Breaking strain"

So, when the buckled girder
Lets down the grinding span,
The blame of loss, or murder,
Is laid upon the man.
Not on the Stuff—the Man!

This was good law when it was written in 1935, and it is still good law in 2021. Will it be good law forever? No one can now say. "The blame of loss, or murder" must in a legal sense be attributed to an independent being, a person.

Currently, although a programmed computer system can often do things that its designer or programmer did not predict, it is not a truly independent being, it is not a moral agent. It cannot form a criminal intent. it cannot really decide to do or not do anything, and it has no internal understanding of what it does, it is not self-conscious.

Presumably the designers of the self-driving car did not intend or foresee that it should hit and kill a pedestrian. Possibly they were negligent, possibly not.

I can imagine a computer system where none of the above statements would be true. Many SF stories have written about such machines. There is dispute among philosophers and AI experts as to whether such a truly sapient computer is even possible. But no one argues that it has already been achieved here on Earth.

When and if a truly sapient computer (or other machine) is achieved, and is recognized as an independent being, many concepts, including legal concepts, will need significant revision. I no longer expect to see that in my lifetime. I am sure it is not yet at hand.

The question asks:

Is there some standard of sentience that the computer would have to achieve in order to indict it?

No specific standard has yet been widely accepted, but if such a situation nears realization there will be. The best current standard seems to be the Turing Test, but thatm has a number of problems, adn i do not think we would be willing to consider a computer system to be a legal person based solely on passing a Turing Test.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.