Following are mainly the problems revolving around conferring AI systems
with legal personality:
It is of prime importance to any legal system because it leads to ascertaining
as to who is capable of acting, who can be bestowed with rights and duties and
thus in this manner, precursor to almost every other issue.
There must be someone in order to be blamed in case things go wrong. There
should be someone who can be sued or having capabilities to enter into contract
etc. In order to deal with liability related issues like automated driving
system entity whose behavior may not be under control of its owner or
There must be someone for purpose of rewarding when things go right like who
would own the intellectual property they create independently.
There could be 2 forms of legal personality:
Natural:Their recognition is guaranteed owing to fact of them being human. There
is a notion in legal parlance that AI could equal humans and that's why a legal
personality as a natural person should be conferred upon them. How the legal
system shall be, whether it should constrain behavior or how it shall shape the
said behaviour for now and when it surpasses humanity for example. Is the
comparison with humans of these AI systems appropriate?
Juridical persons which are non-human entities in their nature and rights and
duties to them are bestowed by virtue of law. Corporations for instance is its
example but other non business entities like religious ones in India can be its
other examples. The question to be addressed is whether assigning a juridical
personality would as a result fill the gaps or prove to be advantageous insofar
as legal system is concerned. Another question is if for instance, some form of
legal personality gets assigned, what content it would have.
Arguments In Favor And Against Conferring Legal Personality To AI Systems
As mentioned above, there can be two categories into which an entity could be
placed into i.e. either Juridical or Natural personality. This section is
categorized into those categories with the former one dealing with theories that
are there concerned to it for justification as to its conferment and then if at
all there arise need to put AI system into that category, what will be content
of it is next sub-heading with feasibility of same ascertained with arguments
for and against the same. Later part deals with Natural personality.
- Juridical Personality
Define of this is given in intro section. Arguments are generally in
relation to nature of liability issue in case as driver-less car or for
trial of criminal robot etc. One needs to delve into various theories in
this direction and whether justifications given for conferring personality
upon corporations could be extended to AI systems too. Several Theories
purported to be in favor are as follows:
Counter arguments against those theories are as follows:
- First theory is the aggregate theory or also known as symbolist theory. It
states that corporation is kind of device formulated by legal system to allow
natural persons in organizing themselves in a group form in order for reflecting
them in terms of their legal relations with other parties.
- Then comes the fiction and the theory of concession, both have separate origins
but ultimately points towards one notion: personality is conferred upon
corporations because a legal system desired to do so. There can be various
reasons to do so, like for encouragement of entrepreneurship or to enhance
stability as to the legal system etc. Therefore, in this manner, the practice of
legislation or judiciary for that matter in recognizing personality can be well
encompassed and extended to AI systems as well.
- The realist theory on contrast holds that corporations are neither fictions nor
just symbols but rather real entities which are preceded by personality
conferment by law. There may be members but their independent actions can be
well attributed to them. Theorists and sociologists tend to regard them as moral
persons too and not just legal. This is in relation to idea that personality is
not just granted but rather deserved. AI systems in the same way deserve to be
recognized as a person.
The theory discussed above namely the aggregate theory has been criticized by
scholars for its applicability to corporations. For it too be argued that it
should be extended to AI systems for same reasons would be an under-statement as
it is least applicable to the AI systems.
The argument that such entities are not just personality owing to the fact of
them being contemplated by law and are also moral persons and that personality
is not just bestowed but deserved is ill-informed in the sense that ultimately,
the term 'person' will stipulate what the law would desire it to stipulate. In
end, its gift of the State insofar as recognition before law is concerned.
The question as to granting personality is not binary in nature but rather of
spectrum nature. Even if it should be granted personality, the grant would
ultimately depend and be guided by rights and duties that would be purported to
be recognized when States would go in direction of making the decision.
- Content of Legal Personality
Recognition as a person before law need not be same for all. Same goes with
rights and obligations that are bestowed. Even in case of natural persons, this
is self reflective from struggles in direction of equal women rights or for that
matter the other marginalized members of society. Therefore, there could be
rights without obligations like when personhood to nature was granted by
Ecuador. Same way, AI can be granted personality where there would be only
Obligations would come into picture in order to address the accountability gaps.
In case of civil liability, the end result is awarding of damages which is in
turn depending upon if wrongdoer is capable of owning property or not. There may
be plausible way out to this like a central fund similar to the regime of
compulsory insurance for addressing the liability question. But in that
scenario, conferring personality would merely be a formality.
So, it becomes important to delve into what shall be the content if such kind of
legal personality is conferred in various areas of law and dealt with below:
In Private law
For: As rightly acknowledged by European Parliament, ability to be sued is one
primal attraction when it comes to conferring personality to AI systems. The
rationale behind is filling accountability gaps.
Against: Situation may arise where the people that currently are bestowed with
responsibility would attempt to do away with that by virtue of escaping from
their liability and shifting same on electronic persons. This is applicable to
corporations too where attempt is made to shield the investors etc. from
liability beyond a said amount of sum and it is justified in lieu of encouraging
investment and entrepreneurship etc. even though there are safeguards formulated
to abuse its form.
Similar kind of mechanism could be employed in case of AI
systems where the said system could be manipulated to protect another person
from liability. The question that arises is that if the person in the first
place is able to do that with a system, does that system indeed worthy of
granting separate personhood?
For: Entering into contracts is another argument, and taking debt and owning
property are its facets. What about accumulation of wealth on part of such
systems? How would they be taxed etc. are some questions put forth in favor of
Against: Usage of electronic agents is not new when it comes to concluding
binding agreements like in case of high frequency trading where it relies on
algorithms concluding agreements with algorithms to conclude agreements on
behalf of natural persons. Autonomy of AI systems could be a challenge to this
in case something goes wrong, but it is resolvable without discourse to a new
As far as taxation is concerned, taxation of robots has been
suggested to be a measure and Bill Gates has suggested that companies that own
them should be taxed but there have been counter arguments saying that it would
lead to negative impact on competition. In future, if need arises, regressive
taxing system on profits can be followed just like the scenario of a company
abusing its market position. Hence, not enough for granting them personality.
What if a head of a company created limited liability company adds an AI system
as member and then resigns, such loopholes might be resorted to by natural
In Criminal Law
For: The ability to be punished is one important argument put-forth in favor of
personality to be assigned to systems.
Against: The rationale behind criminal punishment is retribution,
incapacitation, deterrence, and rehabilitation. For instance, in incapacitation,
it includes exile, amputation etc. and from point of view of corporations,
withdrawing the license or winding up orders. Analogy is made with mad dog or a
faulty vehicle that needs to be decommissioned. But it is more in nature of
administrative than penal and depends on whether guilt can be determined.
or mentally ill considered to be incapacitated can also be detained if they are
deemed danger to themselves or community and thus, do not loose their
personality. Same goes with AI. Measures can be taken like license revocation
etc. without giving personality.
For: Rehabilitation mechanism of punishment which seeks to influence decision to
offend rather than the ability to do so is well suited when it comes to AI
systems with idea that contravention of criminal law to be dealt with by
debugging errors rather than the sins to be penalised.
Against: Legal personality or for that matter, the coercive powers of State are
not at all required in order to make sure that AI systems or Machine Learning do
not lead to such outputs capable of rendering violation of criminal law.
- Natural Personality
The mere fact that one gets born makes him entitled to be person before law.
Although, it was not always like this and there has been history of slaves being
bought and sold like property, or indigenous persons were made comparison with
animals roaming the land etc. Today, anyone who is adult, of sound mind is
regarded as having certain rights and obligations.
It is argued that since AI systems are quickly approaching human related
qualities, that would suffice for them to be recognized before law.
That this is the phenomena of anthropomorphizing AI systems and some
have also formulated a term for it viz. 'android fallacy'. There have been
plenty of experiments showing that people tend to ascribe human qualities post
their experience with humanoid machines like the moral sensibility.
In 2017, European Parliament resolution lead to AI experts across the world to
tender an open letter saying that conferring legal personality to them would be
not apt from perspective of ethics and law. But that argument was on basis that
if such status is granted, it would bring with it all human rights which are
guaranteed under the EU law. This basis itself is reflective of 'android
fallacy'. Another argument against is that in form of 'speciesism' where it is
considered that privilege to human welfare shall be regarded at above level than
that of robots.
That rights should be granted to such systems if not because of
instrumental reasons like that in case of corporation, then for the inherent
reasons like they should be in position to own what they create. Some even
resort to making analogy with darkest history of slavery to support this idea.
The justifications given for making comparison with slavery are of
instrumental nature rather than the inherent reasons to begin with. Moreover,
reliance made upon laws that have been long discarded is not justified. An
alternative can be regarding treatment of animals. They can be bought and sold
but capable of being treated humane.
There was a New York State Court case where
chimpanzees were denied personality rights since they cannot be bestowed with
duties and fails on premise of social contract theory. But the fact is, even
though they cannot be regarded as person but at the same time, cannot be
regarded as a 'thing' merely. Scholars have argued that animals being more close
to humans as compared with AI systems have not been given personhood even though
emotionality aspect is under consideration during that discussion. There are
animal abuse laws and one day there will be some to protect social robots as
well on same line of rational, but whether they will merely meet social
standards rather than biological and technological ones is the question to be
If a human being can be granted ownership for work done by him, why not AI
systems as today, they write reports, compose songs etc. and are worthy of
attracting provisions of copyright law.
The object behind copyright policy is to incentivise innovation.
Generally, there is creative input or intervening on part of human which leads
to result that come out by AI systems. Those cannot be conferred upon system.
But the increasing work of independent nature of AI without human intervention
will make things tough for humans to take credit in long run. In 2020, European
Parliament regarded such works as equivalent to intellectual works in a draft
report but did not felt to make it capable of giving personality of any ind. But
in the end, the human creativity has to be regarded upside that of machine one
in same manner as the kind of 'specieism' that was mentioned earlier.
Most arguments which are in favor of conferring personality to AI systems are
either way too simple or way too complex. There could not be as of now plausible
category into which the personality could be accommodated to the AI systems.
Argument that the explanations for recognition corporations as a juridical
persons from perspective of various theories like symbolist theory, realist
theory etc. could be extended to AI systems too fell short on various grounds.
When one delves into arguments such as need to fill accountability gaps or there
is need for someone to be held liable or so, it was important to ascertain what
might the structure and content of that personality would be. Like for private
law, granting personality could pave way for existing natural persons to escape
liability by shifting blame on AI systems by way of manipulating the same.
regards of imposing damages upon a system, the mechanism such as that akin to
central fund similar to insurance kind regime could be incorporated as owning
property would be requisite for same. That could be done without granting
personality to AI systems. Similar is the situation as far as necessity of
entering into contracts argument is concerned, the utilization of electronic
devices for concluding agreements is not new concept as discussed above and
grievances pertaining to it can be resolved without recognition of them as a
separate new legal entity.
From criminal aspect, legal system can take discourse
to existing methods such as revoking the license or rollback a product to
incapacitate the same in order to prevent further offences and likewise, as
discussed in above sections, there is no need to recognize them as legal
personality or use coercive powers of state to prevent Machine Learning from
committing criminal offences.
Then come arguments which are too indulged into future possibilities and
speculations as opposed to resolving current problems and in that pursuit make
unstated assumptions or suffer from the concept of android fallacy as discussed
above in Natural persons section. For present problems, it would be feasible to
stick to shifting responsibilities of wrongdoing to the users, owners or
manufacturers in place of AI systems themselves and there is no need to
forcefully fit systems into a category on basis of ill-informed assumptions or
future developments till the time these problems that arise can be resolved
through existing legal mechanism.
- Simon Chesterman, Artificial Intelligence And The Limits Of Legal
Personality, 69 International and Comparative Law Quarterly, 819–844 (2020).
- J Dewey, The Historic Background of Corporate Legal Personality, 35 YALE
- G Hallevy, Liability for Crimes Involving Artificial Intelligence
Systems (Springer 2015)
- Y Hu, Robot Criminals, 52 University Of Michigan Journal Of Law Reform
- SM Watson, The Corporate Legal Person, 19 JCLS 137 (2019).
- DA Crane, KD Logue and BC Pilz, A Survey of Legal Issues Arising from
the Deployment of Autonomous and Connected Vehicles, 23 Michigan
Telecommunications And Technology Law Review 256–9(2017).
- S Bayern, The Implications of Modern Business-Entity Law for the
Regulation of Autonomous Systems, 19 Stanford Technology Law Review
- C Mulligan, Revenge Against Robots, 69 SCLRev 579(2018).