white black legal international law journal ISSN: 2581-8503

Peer-Reviewed Journal | Indexed at Manupatra, HeinOnline, Google Scholar & ROAD

Use Of Artificial Intelligence As A Tool To Meet The Ends Of Law Enforcement And Justice By - Dr. Raju Narayana Swamy IAS

AI Enabled  Inventions as IP Owners  : A  Myth or a reality?


Authored By –

Dr. Raju Narayana Swamy IAS

Principal Secretary To Govt Of Kerala



The paper analyses how Artificial Intelligence (AI) enabled systems can be brought into the Intellectual Property (IP) ecosystem. It dwells upon the question of AI- IP interface from three perspectives, viz., (a) AI as a technology to manage IPRs, (b) IP rights as an obstacle to the transparency of AI and, (c) patents as well as copyrights as legal systems that can foster AI. The three step test for obtaining a patent- novelty, inventive step and utility - is looked at through the lens of AI technology. Issues such as patent evergreening, best vs worst embodiment and liability for illegal acts which cannot be traced to human actors are delved into. The article concludes with the need for a uniform treatment of the AI system across the board by bringing in an amendment to TRIPS and the necessity to usher in regulators for adjudication.

Keywords: Artificial Intelligence, Intellectual Property, Super-intelligence, Machine Learning, AI in India, IPR in India.



Artificial Intelligence (AI) is the science and engineering of making shrewd machines. The term was officially instituted by John Mc Carthy (2006) who alongside Marvin Lee Minsky coordinated the Dartmouth gathering in 1956. As per his vision, it is the thought of a program, preparing and following up on data to such an extent that the outcome corresponds to the way in which a smart individual would react in response to a comparable input (Acosta, 2012).In other words, AI is the capacity of a machine to impersonate a canny conduct. Such conduct, it should be referenced here, may  arise either from  a psychological methodology or a computational methodology.


In the Indian setting, the NITI Aayog conversation paper characterizes AI as "a star grouping of advances that empower machines to act with more significant levels of insight and imitate the human capacities of sense, appreciation and activity"(Sony. P & Singh, V. Karthikey, 2019). This arrangement is to be  seen in the context of the educational meaning of AI,as proposed by the research and advisory organization, Gartner Inc., as the technology that turn up to mimic and match human performance typically by acquiring and collating skills, making decisive conclusions, comprehendingmultifarioussubject matters, engaging in communication with the physical environment, strengthening human intellectual and perceptive actionsor supersedinghuman involvementin execution of unusual tasks and missions(as cited in Sony& Singh, 2019).


From Weak AI to Strong AI and Super-intelligence

The contemporary state of AI is prominently alluded to as frail AI. Its qualities are twofold, viz., (a) it has direct human intercession in its creation and, (b) it is restricted to a solitary errand. Siri, for instance, is a frail AI framework utilized by numerous individuals to help them even in daily routine tasks. The following phase of advancement is solid AI wherein human like deduction – with an emotional and cognizant psyche – is prompted. An exemplary model is a "Creativity Machine",commissioned by the US military to plan weapons (Sayler, M. Kelly, 2020). Some information technology researchers are of the assessment that solid AI frameworks may develop to geniuses, outperforming people in the manner of thinking.


The Turing Test

The Turing Test was proposed by Sir Alan Turing (1950), to ascertainif the outcomes being delivered by a machine are the result of its own insight or that of calculations and orders. The test called upon persons to interact with a machine or human and afterwardsto conclude whether they could differentiate if they spoke  to a human or a machine. Turing was of the  view that an AI machine showed insight if the reactions submitted were  indistinguishable  from genuine human reactions. Regardless of the underlying achievement, the test endured turnaround in later years and its application was to a great extent confined to discourse machines and certain testing purposes.


WIPO Classification

The World Intellectual Property Organization (WIPO)propounded three classifications of AI, namely, master frameworks, insight frameworks and common language frameworks. Master frameworks are programs that tackle issues in particular fields of information like diagnosing ailments and suggesting treatment. They depend fundamentally on a hand-made information base and set of rules made by people. However, a framework that  is wholly dependent on  flow of  informationcannot scale and  after a certain stage,  master frameworks got rigid. Also, there are numerous genuine difficulties which are too unpretentious to be in any way addressed by shortsighted intelligent thinking that observes a bunch of rules composed by human specialists. Insight frameworks  enable us  to see the world with the feeling of sight and hearing. Acommon  language program by contrast is planned to comprehend significance of words mulling over various syntactic and literary settings to give a semantic examination.(Kurzweil. R, 1990).


AI vs Machine Learning

Latest advances in AI enable programming of PCs to gain from past experience. An exemplary model is identification of apples from among natural products in a bin of food supplies. By depicting what an apple resembles, we can program a PC so the machine can perceive apples dependent on their shape and shading. AI can be utilized to anticipate whether a client will default on a bank credit or to utilize side effects to foster a clinical determination. Independent driving is another feasible developmentforecastedwith AI. Despite the fact that the numerical thoughts behind AI date back to many years, ongoing advances in information stockpiling, computational speed and sensors have drastically diminished the expense of AI-based forecasts. Obviously, AI is starting to make its presence felt increasingly, more in the everyday setting. An exemplary case is x.ai, a New York City – based startup, that gives a virtual individual associate to book arrangements over email and oversee schedules. A brief framework for understanding AI is illustrated in figure.1


Use Of Artificial Intelligence As A Tool To Meet The Ends Of Law Enforcement And Justice


Authored By - Dr. Raju Narayana Swamy IAS



            With the advent of big data analytics, machine learning and artificial intelligence (AI), the fundamental questions of law enforcement and justice are being reconsidered across the globe. Law is based on two important aspects – predictability and precedence and many are of the opinion that AI can greatly help align these processes. While disagreements are galore as to whether these technologies represent a panacea or whether they will further exacerbate social divisions and endanger fundamental liberties, the two camps agree that the new technologies usher in important consequences. In fact, there are three main ways in which technology is already reshaping the judicial system. First and at the most basic level, technology is assisting to inform, support and advise people involved in the justice system (supportive technology). Second, technology can replace functions and activities that were previously carried out by humans (replacement technology) – the concept of online courts being a classic example. Finally, at a third level, technology can change the way that judges work and provide for very different forms of justice (disruptive technology), particularly where processes change significantly and predictive analytics may reshape the adjudicative role. It is at these second and third levels that issues emerge in terms of the impact of technology on the role and function of a judge. Questions raised in this context include

  1. Can AI enabled programmes extract the accurate position of law from a mass of precedents?
  2. Can robots decide questions of law?
  3. Who should be accountable for semi- automated decisions?
  4. How should responsibility be allocated within the chain of actors when the final decision is facilitated by the use of AI?
  5. Is the “due process of law” denied to the accused when AI systems are used at some stage of the criminal procedure?
  6. Can judgements be replaced by data?

These questions are all the more relevant now that AI has made a lot of inroads within justice systems – in Estonia for adjudicating small claims (robot judges), in China, Russia and Mexico for giving legal advice/approving pensions, in Malaysia towards supporting sentencing decisions, in Austria for sophisticated document management, in Colombia and Argentina for identifying urgent cases within minutes, in Abu Dhabi for predicting probability of settlement and in Singapore for transcribing court hearings in real time- to name a few.


Artificial Intelligence : The Concept

            Among the several definitions of AI, one of the most relevant in the context of justice systems is the one given by the Commissioner for Human Rights:


“An AI system is a machine-based system that makes recommendations, predictions or decisions for a given set of objectives. It does so by


  1. utilizing machine and /or human-based inputs to perceive real and /or virtual environments
  2. abstracting such perceptions into models manually or automatically and
  3. deriving outcomes from these models, whether by human or automated means, in the form of recommendations, predictions or decisions.”

Put in simple terms, AI is a type of computer technology which is concerned with making machines that carry out work in an intelligent way, similar to the way a human would. The technology has evolved from obeying pre-designed and pre-configured codes into a more sophisticated end product, imbued with human-like cognition. AI in order to work needs “big data”. Luc Julia, one of the creators of the digital assistant Siri evokes this image, “if a machine is to be able to recognise a cat with 95% certainty, we need about 1,00,000 pictures of cats.”[i] It needs to be mentioned here that AI comes in many different forms such as speech recognition and image recognition. There are two main strands to current AI technology – one being the rules-based approach and the other based on data analysis. The former underlies many of the document creation products that are used both by lawyers and the lay public, including products such as Legal Zoom. This technology more generally underlies what are called “expert systems” and faces inherent limitations. The latter looks for patterns in large bodies of data and finds relationships and correlations from which it can draw conclusions. This is the kind of AI that underlies products such as translation software and autonomous vehicles.



It needs to be mentioned in this context that the “strong” AIs of science-fiction literature do not exist. This type of AI which would be equipped not only with intelligence but also with conscience remains purely fictional. The machine learning systems currently being developed are described as “weak” AIs and are capable of extracting complex patterns and learning from large volumes of data efficiently and often with high levels of predictive accuracy.


Experiences in other jurisdictions

  1. Estonia

Estonia has announced a plan to delegate some lower value claims to an online court powered exclusively by AI.[ii]The approach includes possible use of a de novo appeal to a human judge.[iii]While providing potentially enforceable actions, this amounts more to a form of AI powered mediation with the litigants free to pursue their legal claims if they are unhappy with the AI generated result. But this may differ from traditional mediation in that simply walking away may not be an option if the AI decree is enforceable – the disappointed litigant will need to invest in and commit to a human-driven judicial process in order to escape the decree.


  1. China

China limits AI to specific kinds of easy cases where the decision parameters are simple and clear.[iv]In Zhejiang province, for example, several thousand dangerous driving and theft cases have been initially decided by AI software subject to review by a human judge. The large sample of similar cases and well –defined determinants of outcome have made this possible. Over time, the level of human review could conceivably be dialled down should experience show that these cases are really as easy as believed. The Chinese experience however points out that even for suggesting outcomes, not all cases are suitable for AI. In order to get to statistically significant results, there must be a large pool of cases with a limited number of factors that can affect outcomes. To put it differently, even in a country as committed to capturing data in its judicial system as China, not all kinds of cases are amenable. In fact, China's ambitious AI system that sought to model intentional murder failed as too few such cases existed to provide an adequate sample and as intentional murder presents in a multiplicity of ways making recognition and characterization more difficult.


China is also a pioneer in using AI to make sure that the resolution of a dispute by a particular court is in line with the results reached by other courts on similar facts and legal issues. Though China is nominally a civil law country, this process has the effect of bringing something like “stare decisis” to Chinese jurisprudence.    In some areas of China, AI robots greet visitors to the court house and help guide them to the appropriate location. There is no reason why this guidance cannot become more sophisticated and helpful over time thereby helping litigants to produce legal forms that are in accord with the requirements of the court and guide them with regard to the court process. Practical examples in this regard include a robot called Xiao Fa which was put in to operation at the law suit center at Beijing No.1 Intermediate People’s Court which can answer questions verbally or take queries on its screen with a key board and the AI enabled robot chatbot Fa Xiaotao using which Wusong Technology is working on digitizing the way courts function.


Mention also needs to be made here of the “smart court navigation system” and “intelligent push system” launched by the Supreme People’s Court in 2018, Beijing’s “rui judge” intelligent research system, Shanghai’s “206” criminal case intelligent auxiliary case system and Hebei’s “smart trial 1.0” trial support system. The smart court SOS in particular is a system powered by machine learning which is being used at various courts that connects to the desk of every functioning judge across the country. It automatically screens out cases for the purpose of reference, recommends relevant laws and regulations, drafts legal documents and alters perceived human errors if any in a court’s judgement. Moreover through facial recognition techniques, Chinese officials have evolved systems to figure out suspects from an ocean of individuals inside an arena which without man-made intelligence innovation would have been impossible. China is also experimenting automation in prisons. A jail that houses some of China’s most high profile criminals is reportedly installing an AI network that will be able to recognize and track every prisoner round the clock and alert guards if anything seems out of place.


  1. USA

Judiciary in the US has already started using an AI programme called Public Safety Assessment (PSA) before deciding whether or not an accused should be released on bail.[v] This software calculates the risk of recidivism and flight ( ie) risk of defendant again committing the crime and risk of his escape from the clutches of law. The AI software calculates the risk score by taking into account several factors such as

  1. whether the current offence is violent
  2. whether the person had a pending charge at the time of the current  offence
  3. whether the person has a prior misdemeanour/felony conviction
  4. person’s age at the time of arrest
  5. how many times the person failed to appear at a pre-trial hearing in the last two years

Other factors which can assist judges in arriving at bail, parole and probation decisions can also be incorporated into the algorithm.


            Mention must also be made of the SSL (Strategic Subject List) introduced in Chicago to predict those individuals who are likely to be involved in gun violence. US also makes use of “e Discovery” – an automated investigation of electronic information before the start of a court procedure.  e Discovery relies on machine learning. A more controversial tool COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) has been used to assess recidivism risk. COMPAS takes in 137 items of information and comes up with a risk score from 1 to 10. However, it has been criticized as biased against black defendants. In fact, the NGO Pro Publica analysed COMPAS assessments and published an investigative report arguing that the algorithm was racially biased.


No discussion on COMPAS will be complete without a reference to the judgement in Loomis vs Wisconsin (2016). The algorithm identified Loomis as an individual who presented a high risk to society and the first instance Court decided to refuse his request to be released on parole. In the appeal, the Supreme Court of Wisconsin decided that the recommendation from the COMPAS algorithm was not the sole ground for refusing his request to be released on parole and hence the decision of the Court did not violate Loomi’s due process right. The Court was in fact neglecting the strength of the automation bias. By claiming that the lower Court had the possibility to depart from the proposed algorithmic risk assessment, the Court ignored the social psychology and human-computer interaction research on the biases involved in all algorithmic decision making systems which shows that once a hi-tech tool offers a recommendation, it becomes extremely burdensome for a human decision maker to refute such a recommendation.


  1. UK

A tool called HART (Harm Assessment Risk Tool) has been used by the UK to forecast which criminals are most likely to reoffend and suggest what kind of supervision a defendant should receive in prison.[vi] The tool uses random forest forecasting which is a machine learning technique. HART was developed in collaboration with academicians at Cambridge and is built on five years’ worth of data on people. It makes predictions on the basis of 33 different metrics, 29 of which relate to past criminal history and the remainder to demographic data such as the individual’s age, gender and postcode[vii]The usage of postcodes  as a metric of analysis has garnered repeated criticism of this tool as critics argue that such a situation amplifies existing patterns of offending.[viii]The application has also shown clear difference in opinion between human and algorithmic forecasts. In fact, comparison in the use of the algorithm found that the model and officers agree only 56.2% of the time.


  1. Brazil

              Brazil uses an AI tool called VICTOR to conduct preliminary case analysis to reduce the burden on the Court.[ix] The tool supports the Brazilian Supreme Court by providing analysis of cases that reach the Court using document analysis and natural language processing tools. The goal of this tool is to accurately and quickly track resources that deal with issues of “general repercussions. “ This concept is intended to ensure that only questions that are truly relevant to the wider society are heard by the Court. To put it a bit differently, appeals that reflect only the unsuccessful party’s unwillingness to accept defeat are excluded.


            VICTOR has been highly beneficial for courts in Brazil. Earlier, the exercise was conducted by civil servants based on binding precedents and would take about 40 minutes for each law suit. VICTOR can do this exercise in five minutes.


            However, a criticism in the pipeline is that appellants are not informed when VICTOR is used as its pilot version randomly picks up appeals to evaluate. This is in possible violation of the Brazilian Data Protection Law which demands that decision making should be fair, transparent and informed.


  1. A few other countries

                  In the Middle East , in collaboration with the private sector, Abu Dhabi Judicial Department as a part of their “Justice Intelligence” Project has been using technology to predict the possibility of settlement of cases.[x]The tools that are being used can predict the probability of settlement by upto 94% of the time. In Argentina and Columbia, a tool called Prometea has been used by the Public Prosecutor’s Office of Buenos Aires and the Constitutional Court of Columbia respectively to identify urgent cases in just two minutes which would normally take a human being 96 days.[xi]In Singapore, a speech translation system has been deployed by Courts. The tool utilizes neural networks trained with language models and domain- specific terms to transcribe court hearings in real time, thus allowing judges and parties to review oral testimonies in Court instantaneously. [xii]In countries like Russia and Mexico, robots are providing services like legal advice to citizens and aiding judges to identify if pensions should be granted. In Austria, AI is being used for sophisticated document management and as a digitization assistant of existing analogue files. In Malaysia, AI is being used to support sentencing decisions.[xiii]


The Indian Saga

            AI in the Indian judiciary is still in its nascent stages of development. The advent of AI within Indian Courts was initiated on 26th November 2019 with the launch of a neutral translation tool called SUVAAS (Supreme Court Vidhik Anuvaad Software) which has been trained using machine learning processes. It has the capability of translating English judgements and daily orders into nine vernacular scripts. The Supreme Court’s AI Committee is also working on a composite new tool named SUPACE (Supreme Court Portal for Assistance in Court Efficiency) which will target different processes like data mining, legal research, projecting case progress etc.


            The AI-powered work flow of SUPACE has four parts:

  1. File Preview

The case files, typically available as PDFs, can also be converted into text. There is also a search tool to browse through all files


  1. Chatbot

The text and voice enabled chatbot helps to give a quick overview of the case in a matter of minutes by answering simple questions such as “ What is the matter about?” or “Which fundamental rights of the petitioner are violated?”. Chatbot can switch between documents to fetch the right answer while allowing the user to check the source of the answer. This bot suggest further questions to be asked for better understanding and the entire question summary can be printed by the user.


  1. Logic Gate

This fact extraction system of the chatbot is divided into four parts : Synopsis, FAQs, Evidence and Case Law. These give information about the case such as overview, chronology, judgement and so on. With enough training and refinement of the algorithm, there will come a time when any and every question, factual or contextual, will be answered by the chatbot.


  1. Notebook

This is the integrated word processor which truly makes the tool an end –to- end system. A brief summary of the case can be prepared by collating all information auto-extracted from the database using the AI .In addition, voice dictation can be used to prepare notes on this comprehensive drafting tool. Therefore without typing a word, a summary document is ready as a soft or hard copy.


On the research front, IIT Kharagpur has released an AI- based method to automate reading of legal case judgements. It is based on deep neural models to understand the rhetorical role of sentences. It also needs to be mentioned here that the pandemic has led to a surge in discussion around increasing digitization through the e Courts Project, creation of virtual courts and the potential of Online Dispute Resolution (ODR) and that within this conversation, AI has become an increasing talking point. However, a word of caution is needed here – while pilots are promising, there is a need to identify steps for scaling these technologies and their adoption.


Is AI a Panacea?

            The question revolves around a fundamental issue: Will we as a society ever be willing to delegate fundamental rule - making powers and assign assertion of the legitimacy of the state to non-human entities ? Only by matching the real potential of AI with the full range of judicial functions can we give a non-hyperbolic assessment. The reality is that the current capability of AI is limited to specialised tasks and the roles of judges are so generalised that there is no near-term possibility of AI wholly and satisfactorily displacing judges in high stake cases. The function of AI with the best evidence so far of success is structuring large amounts of information which could make administration of justice more efficient. In other words, AI can be used partially – where human discretion may not be needed. An evolutionary step forward can be advancement in predictive big data policing instruments. There can be a two pronged strategy behind this. First, advancements in AI promise to make sense of enormous amounts of data and to extract meaning from scattered data sets. Second, they can be aimed at regulation of society at large and not just the fight against the crime. A classic example is ‘function-creep’ , Singapore’s total information awareness system programme.


            AI tools can be used to penetrate deeply into the preparatory phase of the crime which is yet to be committed as well as to scrutinize already committed crimes. A distinction is however to be made between tools focusing on risky individuals (heat lists) and tools focusing on risky places (hot spots). A leaf can perhaps be drawn out of the experience of the Dutch children’s rights organization Terre des Hommes which was the first NGO to combat webcam child sex tourism by using a virtual character called Sweetie.



            AI is not a kind of magic; it is technology with the capabilities and limitations inherent therein. The intelligence of AI is limited by the design and input of human programs. An excessive reliance on such systems may result in legal issues being decided by computer programs. The fact is that “algorithmic partiality” is entrenched in the method AI algorithms work.[xiv]For example in 2015, Google’s photograph application inaccurately labelled a photo of two dark individuals as gorillas since its calculation had not been prepared with suitable pictures of people of dark skin. In another matter, the AI judged beauty contest for the most part picked white members as victors since its training was done on pictures of white individuals. [xv] These tools have the capacity to become prescriptive, potentially over shadowing case specific reasoning and reduce judicial decisions to purely statistical and algorithmic outcomes. In fact, there are two significant issues with a pervasive integration of AI within the justice system – value lock-ins which can stagnate legal and jurisprudential evolution and the alteration of the judiciary’s constitutional role under the doctrine of separation of powers. The impact of AI systems may have distorting effects on the fundamental cornerstones and architecture of liberal democracies including the limitation of political power by the rule of law. Human rights may also be impacted through the use of automated processing techniques and algorithms. To be more specific, in an AI –centric judiciary while fostering the rule of stare decisis, it is a plausible fall out that precedents become stagnant. Moreover as Yutaka Matsuo, a Japanese scholar pointed out, civil litigation especially divorce or property inheritance disputes may be better handled by people because it involves a lot of emotional factors.


The Road Ahead

            The role of a judge is a complex one. It can incorporate activism, interactions with people, dispute settlement, case management, public and specific education activities, social commentary as well as adjudicatory functions that may be conducted with other judges or less commonly in some jurisdiction with lay people. Given this variation, it is difficult to determine how developments in AI may reshape the judicial role. Human judges are temperature-saavy and as participants in social life, they have the same ability to empathize with the general public and are committed to achieving empathetic justice. To put it a bit differently, against the backdrop of AI’s full penetration into the judiciary, the concern is not that “machines are starting to think like humans, but that humans are beginning to lose their unique ability to think and become vassals of machines.” Everyone and every case is unique. Each requires human judgement and the vital and very natural ability to empathy that AI systems cannot provide. Although AI will play a key role in the trial process in the years to come, it will be in a subordinate position – only assisting the judge to handle the case – the judge continuing to be the key and core of the trial work. Thus co-bots rather that robots will play a role when we intend to apply AI to the field of law. The road ahead is perhaps one of utilizing AI in addition to people – as airplane pilots use autopilot.

            Technological advances in the form of AI based programs will benefit the judges with increase in memory, enhancement in the ability to manage and process information as well as reduction in occurrence of fatigue. In international arbitration, the use of AI has been predicted for a variety of tasks, including appointment of arbitrators, legal research, drafting and proof-reading of written submissions, translation of documents, case management and document organization, cost estimations, hearing arrangements (such as transcripts or simultaneous foreign language interpretation) and drafting of standard sections of awards. There are several processes in the administration of law that are repetitive and mechanical (such as scheduling of hearings and e-filing) and the use of machine learning in these mechanisms can greatly enhance efficiency and improve case flow management. Decision making in areas like rate of taxation can be covered by AI and thus technology can play a significant role in tribunals like ITAT (Income Tax Appellate Tribunal). AI models like computer vision and natural language processing can be used to generate statements and evidence documents.  Leveraging such technology will help to reduce human dependency and errors in information input. To put it a bit differently, the initial focus of use of AI when it comes to improve the work of judges must be on document/case management and research.


            In fact, the main categories where AI can be used at present are:-

  1. Advanced case – law  search engines
  2. Online Dispute Resolution
  3. Assistance in drafting deeds
  4. Analysis (predictive, scales)
  5. Categorization of contracts according to different criteria and detection of divergent or incompatible contractual clauses.
  6. Chatbots to inform litigants or support them in their legal proceedings.

The following tabulation speaks volumes for itself.

Application and Examples of AI in Legal Sector[xvi]




Legal Application



Document Drafting

Drafting contracts, form filling using chatbots

Legal Zoom


Contract Review and Management

Identify issues/risks

Provide standard clauses when drafting


Kira Systems

Law Geeks


KM Standards

Document Management

Storing and easy retrieval, auto template creation and scanning documents using OCR

Docubot by 1Law

E-discovery/Document Review

Search for necessary (other) facts from internet for analysis and decision. Use key words.

Predictive coding



Due diligence

Review background information and prior cases. Highlight and classify essential clauses

Kira Systems

Legal Research

Find arguments and reasoning reported in the past for assessing similar arguments

Ross Intelligence

Fast Case

Thompson Reuters


Smart contract

Provides an easy way to reference and trigger an Ethereum – based smart contract to manage contractual promises

Open Law


Over a period of time, these uses can lead to more sophisticated ones that require considerable methodological precautions as well as uses to be considered following additional in depth scientific studies. A day may also come when AI can be put to use successfully in areas considered today with most extreme reservations- use of unbiased algorithms in criminal matters in order to profile individuals being a classic example.

At one level, AI is an alien form of intelligence and will be – even if it achieves generalized capabilities – no more like humans than reptilian visitors from another galaxy would be. To have such an intelligence create and extend laws, despite being so far removed from being a member of the body politic, comes up against the legitimacy of the judicial system. Whether our societies are ready to accept that involves issues far beyond technological capability. This is especially true in today’s milieu wherein the neutrality of AI remains constrained and where justice is like the top of a mast, swinging violently at the slightest movement of the hull.



  1. Zichun Xu (2022) Human Judges in the Era of Artificial Intelligence: Challenges and Opportunities,  Applied Artificial Intelligence, 36:1,2013652.


  1. Christopher Rigano (2019) Using Artificial Intelligence to Address Criminal Justice Needs, NIJ Journal 280


  1. Mvea U (2018), Japan Considers Crime Prediction System using big data and AI, Japan Times, 24 June 2018 (Online)


  1. Zavrsnik A (2019), Algorithmic justice: algorithms and big data in criminal justice settings, Ekr.J.Criminol.


  1. Dewan S (2015) Judges Replacing Conjecture with Formula For Bail, The New York Times


  1. Ferguson AG (2017), The Rise of Big Data Policing : Surveillance, Race and the Future of Law Enforcement, NYU Press, New York


  1. Freeman K (2016), Algorithmic Injustice : How the Wisconsin Supreme Court failed to protect due process rights in State Vs Loomis, NCJ.Law Technol 18(5)


  1. Tania Sourdin and Archie Zariski (eds)(2018), The Responsive Judge : International Perspectives, Springer Nature
  2. Aryan A (2019), Law firms take baby steps in AI at increase efficiency and cut costs http://www.nishithdesai.com


  1. Branting K (2003), An Agenda for Empirical Research in AI and Law in evaluation of Legal Reasoning and Problem Solving Systems


  1. Hali AC (2018), How law firms can benefit from Artificial Intelligence, https://www.lawtechnologytoday.org/2018/11/how-law-firms-can-benefit-from-artificial-intelligence/


  1. Hubert Dreyfus (1992) What Computers Still Can’t Do : A Critique of Artificial Reason, Cambridge, The MIT Press.




[i] L.Julia,L’intelligence artificielle n’existe pas, First Edition, Paris 2019

[ii] Erie Niiler, “Can AI be a Fair Judge in Court ? Estonia Thinks So “, WIRED (March 25, 2019,7.00 AM,https://www.wired.com/story/can-ai-be-fair-judge-court-estonia-thinks-so)

[iii] Tara Vasdani, From Estonian AI Judges to Robot Mediators in Canada,  UK, The Lawyer’s Daily, https://www.lexisnexis.ca/en-ca/ihc/2019/06/from-estonian-ai-judges-to-robot-mediators-in-canada-uk.page

[iv] Jinting Deng, ”Should the Common Law System welcome Artificial Intelligence : A Case Study of China’s Same – type Case Reference System” 3 GEO.L.TECH REV 223 (2019)

[v] Arnold Foundation Launches Expansion of Public Safety Assessment Tool https://the crimereport.org/2018/04/25/arnold-foundation-launches-expansion-of-public-safety-assessment-tool/

[vi] Marion Oswald and others “Algorithmic risk assessment policing models: lessons from the Durham HART model and Experimental proportionality” (2018)2 Information & Communication Technology Law

[vii] Patricia Nilsson, “UK police test if computer can predict criminal behaviour”, Financial Times 6th February 2019.

[viii] Alexander Babuta, Marion Oswald and Christine Rinik, “Machine Learning Algorithms and Police Decision-Making: Legal, Ethical and Regulatory Challenges” (Royal United Services Institute for Defence and Security Studies 2018.

[ix] Daniel Becker and Isabela Ferrari, “ VICTOR: the Brazilian Supreme Court’s Artificial Intelligence: a beauty or a beast? “ https:// sifocc.org/app/uploads/2020/06/ Victor- Beauty- or-the- Beast.pdf

[x] ADJD leverages judicial insights to transform the justice delivery process (SAS Institute)www.sas.com/en-ae/customers/adjd-judicial.html

[xi] Juan Corvalan and Enzo Cervini, Prometea experience: Using AI to optimize public institutions (Cendap, 1st May 2020) and Irma Isabel Rivers, the implementation of new technologies under Columbian law and incorporation of artificial intelligence in judicial proceedings ( International Bar Association, 5 November 2020)

[xii] Michelle Chiang “State Courts and A STAR’s Institute for Infocomm Research Collaborate to Develop Real-time Speech Transcription System for Use in Courts”(State Courts, Singapore, 14 December 2017)

[xiii] Olivia Miwil, “Malaysian judiciary makes history, uses AI in sentencing”, New Straits Times, 19 February 2020.

[xiv] Dickson, B (2018)”What is algorithmic bias?” Available at https://bdtechtalks.com/2010102126/racist-ai-deep-learning-algorithms/

[xv] Artificial Intelligence beauty contest does not like black people, The Guardian, 8 September 2016

[xvi] Geethanjali Chandra, Ruchika Gupta and Nidhi Agarwal (2020), Role of Artificial Intelligence in Transforming the Justice Delivery System in COVID-19 Pandemic, International Journal on Emerging Technologies,11(3)


Fig.1. Framework for understanding artificial intelligence(Source: Russel and Novig, 2009 as cited in Kathuria et.al, 2020)

Machine Learning is a subset of AI. The concept discusses the ability of machines to solve problems by learning through the data, beyond the programming and it involves three characters, namely, supervised learning, unsupervised learning and reinforced learning. Figure.2 illustrates the characteristics of AI and ML.

Description: Description: A picture containing text, electronics

Description automatically generated

Fig.2. Machine Learning as a subset of AI.

It is pertinent here to highlight the following concepts of deep realizing, reinforced learning and transfer discovering. Deep realizing mirrors the action in the layers of neurons in the cerebrum to figure out how to perceive complex examples in information. This  maybe the most encouraging innovation where neural organizations are prepared on very huge informational collections. Reinforcement learningrelates to the programming specialists that learn objectiveoriented conduct by experimentation in a climate that gives prizes or punishments by accomplishing that objective. And the transfer discoveringrefers to that centers around  utilizing information acquired in one issue to an alternate or related issue.


The AI-IPR Intersection

The AI-IPR convergence can be comprehensively arranged under three heads:


i.AI as an innovation to oversee IPRs

Across the globe, IP workplaces have conveyed different AI applications, exemplary models being WIPO Translate and WIPO Brand Image Search that utilize such applications for computerized interpretation and picture  recognition. Notice should be made here of the 2018 gathering coordinated by WIPO to examine these applications and energize their sharing.


ii.IP rights as an impediment to the straightforwardness of AI frameworks

In a period of straightforwardness and responsibility, an inquiry emerges regarding whether this necessity will keep on being fulfilled in cases wherein the AI cycle includes components that are misty for legitimate or mechanical reasons (Wexler, R., 2018).Indeed IP rights as a rule and proprietary innovations specifically could make hindrances and raise a contention between IP arrangements from one viewpoint and the social need of straightforwardness. The  need of the hour lies in featuring the truth that revelation for fulfilling these objectives does not concern the algorithmic guidelines, but just their outcomes.


iii.IP as a legal system that can protect, nay foster AI.

Patent and copyright are the most pertinent frameworks of assurance with respect to AI. In any case, when patent laws were imagined, the idea of machine as a creator didn't exist. Along these lines patent laws overall presented innovation rights just to people.An example for this is the Japanese law specification implying that  only an individual can be a creator, and not any machines. What's more, the circumstance has not gone through  a sea  change even today.To refer to a model, an AI framework dedicated as DABUS was named as the creator in patent applications documented in UK, US and Europe in 2017 (Ireland &Lohr., 2020).But the equivalent was dismissed in  all  the three because of it not being a legitimate individual.In this manner, from a patent point of view, the accompanying issues need extraordinary notice:


a.Whether AI as a development is qualified topic. (In many  countries, calculations without help from anyone else qualify as dubious frameworks lacking specialized character and consequently cannot be ensured protection under IP laws. Nonetheless, it will be counterproductive in the event that we adopt a sweeping strategy that  patents ought not be granted to AI-based creations )


b. Who is the valid and first innovator? (Should the law allow that the AI application be named as the creator or would it be advisable for it to be indicated that a person be named as the designer? Provided that this is true, should the law let the partners take the choice by interior courses of action with regards to how the human designer is to be resolved. In the event that we award patent to AI as designer, would it be able to be relegated to the gathering who will get most extreme benefit through commercialization? Would ownership be able to be chosen based on Coase Theorem?)


c.Who claims and is in this way obligated for the demonstrations of the AI innovation? Does the legitimate duty of the illicit activity of an AI lie with its proprietor or its client or its administrator? Should the position of the maker being at risk regardless of him lackingmensrea or even actus reus go through  a radical change? In the event that the reason for the illicit demonstration cannot be  attributed to a human entertainer, who has the risk?

d.Interpretation of non-obviousness


e.Issues relating to divulgence (explicitly how it very well may be satisfied where calculations of AI are not static but rather change over the long haul and handling best versus most exceedingly awful encapsulation issues: AI application maykeep the best exemplification undisclosed and get patent without total honesty.)


f. The manner in which any harms be resolved in case the AI copies a creation or replicates an innovation.


g.Adequateness of current laws. Should a sui generis arrangement of IP rights for AI produced creations be raised? Or on the other hand should the AI-IPR interface be required to be postponed till the D-day shows up when AI innovation is better perceived? (considering the way that at the current phase of improvement, instances created by totally self-sufficient AI frameworks are rare).


The Indian Context: Patent Protection for AIs

Subject matter eligibility

Artificial Intelligence empowered frameworks can make innovations which commonly result  from the use of human intellectual cycles. Nonetheless, there are legal hitches. For example, the disallowance in Section 3(k) of the Indian Patents Act 1970 (as altered in 2002)has triggered a hornet’s nest and has led  to patents  at times being allowed to mixes of equipment and programming or programming with certifiable specialized applications.The Indian Patent Office's position on patentability of PC related developments needs clearness. Anyway a silver line is the  elimination  of the inflexible prerequisite of just programs related to a novel equipment being qualified for a patent . However, we have a long way to go. The need of great importance is a strong system for protecting AI developments, the sign of which will be consistency, consistently guaranteeing that the country stays responsive towards trend-setters. Dismissing all AI  patents on the reason that all AI will utilize the fundamental  modalities  of information assortment, normalization, re-repeat/self-AI, information association, information handling  and output as wanted by human cerebrum will be counterproductive. One can  rely on  the experience of the European Patent Office which has effectively held its first gathering on AI and  patenting. Obviously, the  focus of such a system will be to make India a maker of AI instead of a uninvolved adopter of the equivalent. Simultaneously, AI which can be a potential danger to mankind maybe sorted out as "destructible/perilous development".

Who can apply for a patent?

As innovation pushes ahead from a period of  frail  AI to solid AI, also of  superintelligence,  the  query that  evolves is whether AI innovations can be considered as  creators. Be that as it may, this is as yet an ill-defined situation.  Section 6 of the Patents Act, 1970 endorses that any individual professing to be the valid and first designer of the creation can apply for a patent. This expression is characterized in Section 2(1)(y) as follows:. It does exclude either the initial shipper of an innovation into India or an individual to whom a creation is first imparted outside India.


The part advances an exclusionary definition and doesn't explicitly express that the valid and first innovator ought to be a human.  Thus the Act gives the fortitude to incorporation of works by AI frameworks. In any case, the drawing on the divider isn't so clear. For example, Section 2(1) (p) characterizes the expression "patentee" as an individual for the time being entered on the register as the grantee or owner of the patent (Indian Patent Act, 1970). The Act  also talks about  individuals occupied with or advancing exploration in the very field as that to which the creation relates.


The above  discussion  expresses the view  that it ought to be an individual ( legal individual) and hence the aim of the governing  framework for the Act overall can be perceived to be shifted  towards entities which are persons in the eyes of law. This  underlines the need to correct the enactment to suit the  evolving scenario  of advancing logical frameworks.


The Three Step Test

As regards innovations by AI empowered frameworks, the greatest test towards acquiring a patent is fulfilling the three stage test . The term 'new' is not characterized in the Act. In spite of the fact that the expression "new  invention " is characterized in the Act, this definition is  superfluous  as the term is not utilized elsewhere in the Act. Consequently depending on the precedent-based law significance of the term, we can securely infer that a case is viewed as new if every one of the components of the case cannot be found in a solitary  prior art,  whichhere implies everything made accessible to general society through a composed or oral depiction, by use or in some other  way before the date of development of the invention (Glaverbel SA vs. Dave Rose and Others, 2010 (43) PTC 630). The essential  query  that emerges with regard to AI can thus be expressed in the accompanying terms.


Tripathi and Ghatak (2018), posed a question, "While an AI framework will unquestionably  draw upon earlier craftsmanship, because of its administering human researchers taking care of input data, is it genuinely competent to  arrive at a judgment on whether its innovation can represent something novel?"


The  aspect of  inventive step is  more confounded. The  Act defines the term  under Section  2(ja).

The Supreme Court in the Novartis case separated Section 2(ja) into its components in the following way:


"It [The product] should appear because of a development which has an element that: (a) entails specialized development over existing information, or (b) has a monetary importance and furthermore, (c) makes the creation not clear to an individual gifted in the craftsmanship".


In the light of the abovementioned facts, it should be  stated  with regard to AIs that odds of making developments on existing models or ideas which  are not clear to  individuals talented in the workmanship is surely more hard to accomplish than  mere novelty. Obviously, the innovation should initially progress to furnish these frameworks with a human-like insight so that careful decisions in the new circumstances can be made by them(Trpathi&Ghatak, 2018)


Issues identified with  evergreening:

A significant inquiry that should be  looked into is whether an AI patent application referring to another AI application will  make  the very nature of the creation crumble as even a minor intelligent change would prompt another development.Whether in such cases we need to  bring in legal provisions akin to Section 3(d) of the Patent Act is an issue that should be debated exhaustively. Obviously, this point is significant not  just to keep away from patent evergreening  but in addition to manage the issue of patent  trolls.


Issues relating to  provisional applications:

The approaches towards the manner in which thetemporary applications need to be permitted (as simple  expression of thought to guarantee  priority date will give a  timeline of one year to widen  claims to a limitless degree) should be examined.





Copyright and AI

The primary inquiry brought up in this setting is whether copyright ought to be credited to unique scholarly and imaginative works that are in self-governing mode, produced by AI or should a human maker be required. It should be referenced here that even craftsmanship of Picasso have been reproduced by AIbased frameworks and in 2018, one such work was sold for close to half a million US dollars(Emerging Technology from the arXiv, 2019). In any case, the reality stays that AI workmanship is a subset of generative craftsmanship and is algorithmic - repeatable in nature to be explicit - and regularly open source - shareability being its trademark. One side contends  that systems cannot be as  inventive as humans while the other  argue it contrary (Gelender, 1994). The most acknowledged answer as on date is that, while AI applications are fit for delivering such works in self-governing mode, this limit does not fit with the copyright framework which is  after all  connected with the human inventive soul. The practitioners consider the Lovelace test to be better than the Turing test. Hypothetically, Lovelace states the viewpoint that machines  do not possess  inventiveness (Neill, Sean O, 2014), highlighting the rationale that inventiveness is the capacity to do the eccentric, dissimilar to something machines consistently do. Machines, the corridor sign of which, is rule bound to conduct (and hence AIs) cannot be brought within the ambit of copyright  framework . The counter view depends on decisions that the nonhuman idea of the wellspring of a work ought not be a bar to copyright.


The US copyright office's update to the Compendium of Practices (December, 2014) adds weight to the  first view.Nonetheless, theIPClause of the US Constitution does not unequivocally specify a human necessity. Notice additionally should be made of the WIPO meaning of IP that  dwells on  manifestations of the psyche yet does not determine whether it should be a human imaginative mind.Notwithstanding these, as of late a San Francisco Court  held that creatures not being people  do not have  locus standi under Copyright Act to sue for  infringement (Naruto Vs. Slater, No. 16-15469, 9th Cir. 2018). Obviously, the judgment built up contentions that if Naruto, the monkey cannot sue for copyrightviolation, comparable ought to be the circumstanceforAIframeworks. Also, as appropriately brought up by Tripathi and Ghatak(2018) regardless of whether nations confessed to giving copyrights to works crafted by an AI, the topic of who gets that copyright stays mysterious in the light of the fact that the current status of law requires a legitimate personhood of a holder, something which an AI needs except if its maker is conceded that for its sake. Nonetheless, what occurs if the AI framework was  based on sale-purchase  remains an open question. The appropriate response lies for the maker in nations like England and New Zealand, however this actually does not address the above question in its totality.

Three milestone decisions need references here:

a.Burrow Gilles Lithographic Co. v Sarony (III US 53 (1884))

The case talked about the chance of giving copyright assurance to an item which is the yield of a machine. The Court held that absolutely mechanical work is essentially not imaginative.  If methodology on these lines is  followed , allowing copyright for works made by AI would be troublesome.


b. Bleistein v Donaldson Lithographing Co, 188 US 239 (1903)

 Justice Holmes depicted the uniqueness of human character and specified  it to  be  essential to get a copyright.


c.  Alfred Bell and Co v Catalda Fine Arts Inc. 191 F. 2d 99 (2d Cir, 1951)

The Court brought down the norm for  originality and held that for the work to be  so , it should not be a duplicated one. This judgment was a  relief for the promoters of copyrights for AI created  works as it is not  replicated despite the fact that it is produced through calculations.


In the Indian setting, the test to copyright for  works of AI is Section 2(d) of the Copyright Act, 1957 which  defines author as a person.


For an individual to make a work, nearness of the individual with the work is significant and thus  person  here implies a human or a  legal  individual,  a lot to the dismay of promoters of copyright to AI frameworks.



As we move away from IA (Intelligent Automation) to AI driven by machines, the inquiries around ramifications of such an innovation are developing. Daimler-Benz has effectively tried self-driving trucks on open streets, AI innovation has been applied successfully in clinical headways, a film composed by an AI appeared online as of late and AI has even discovered its way into the  advocate fraternity. Also, Sophia, a social humanoid robot created by Hanson Robotics, a Hong Kong based organization as of late turned into a citizen. Another energizing field interweaved with AI is the idea of Artificial Neural Networks (ANNs)  frameworks of equipment and programming designed after the activity of neurons in the, cerebrum. Neural  nets are viewed as venturing stones in the quest for AI. The principal computational model of ANNs - prevalently called  threshold logic - was created by Warren Mc Culloch and Walter Pitts in 1943 (Palm, G., 1986).From there on ANNs have progressed significantly especially due to their particular capacity to distinguish the fundamental connection between various arrangements of information and because of their dynamic nature - adjusting to changes in yield so that they give the best achievable outcome without changing  the input nodes. A significant capability of ANNs lies in the fiscal field. Anyway exactness of ANNs relies upon the design chosen for a particular issue and  training pattern of ANNs, among different elements.


Add to this, the issue of "deep fakes" and the situation is much more intricate. These are AI-improved phony pictures and recordings that take influence of an AI calculation to embed faces and voices into video and sound chronicles of real individuals and empowers the making of  impersonations wrongly depicting individuals saying or doing things they never said or did. In 2012, an AI chatbot named Sim Simi purportedly figured out how to show itself 'Thai' through correspondence with clients in Thailand. Utilizing the new dialect and expressions, it had gained from dealings with clients, Sim Simi went on purportedly to slander the Thai Prime Minister (Metaratings, 2012)


In this unfurling situation, what is required is a uniform treatment of the AI framework in all cases wherein countries who are signatories to multilateraltrading arrangements start to perceive its  presence  by getting through a revision to TRIPS. Passing of an AI Information Insurance Act which could introduce the institution of a controller to settle  and adjudicate acts of AIs and all the more explicitly set forth solutions for common and criminal offences carried out by  them  is additionally the need of great importance. It should  bring in laws to keep honest makers from being indicted for demonstrations of the AI for which they have no control what so ever. Also, these actions should introduce visionary advances focused at determining how solid AI and  superintelligence  ought to be treated in the IP system. For, we must be ready for the D-day when machines implement, safeguard and even indict. Obviously, the test before the comity of countries and its inhabitants is to bridle this stunning innovation for the advancement of humankind by establishing the framework of a strong legitimate system, nay an AI explicit, yet humankind driven law.



Alfred Bell and Co vs  Catalda Fine Arts Inc., 191 F. 2d 99 (2d Cir, 1951) https://law.justia.com/cases/federal/appellate-courts/F2/191/99/91570/

Bleistein vs  Donaldson Lithographing Co., 188 US 239 (1903) https://supreme.justia.com/cases/federal/us/188/239/

Burrow Gilles Lithographic Co. vs  Sarony., (III US 53 (1884) https://supreme.justia.com/cases/federal/us/111/53/

Emerging Technology from the arXiv (2019), MIT Technology Review, https://www.technologyreview.com/2019/09/20/132929/this-picasso-painting-had-never-been-seen-before-until-a-neural-network-painted-it/

Glaverbel SA vs. Dave Rose and Others, 2010 (43) PTC 630, https://vlex.in/vid/ glaverbel-s-vs-dave-572132818.

Indian Patent Act (1970).,available at https://ipindia.gov.in/writereaddata/ Portal/IPOAct/1_31_1_patent-act-1970-11march2015.pdf.

Ireland, Imogen. &Lohr, Jason (2020), DABUS’: the AI topic that patent lawyers should be monitoring, ManagingIP., https://www.managingip.com/ article/b1n8q624s4vyv4/dabus-the-ai-topic-that-patent-lawyers-should-be-monitoring.

Kathuria, R.,Kedia, M. &Kapilavai, S. (2020). Implications of AI on the Indian Economy.  Indian Council for Research on International Economic Relations, New Delhi.

Mccarthy, J. & Minsky, M. & Rochester, N. & Shannon, C.E.. (2006). A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence. AI Magazine. 27.

Metaratings,  ( 2017, Feb.07), SimSimi chatbot banned in Thailand, Telecom Asia, https://www.telecomasia.net/blog/content/simsimi-chatbot-banned-thailand/  

Naruto Vs. Slater, No. 16-15469, 9th Cir. 2018,  https://law.justia.com/cases/ federal/appellate-courts/ca9/16-15469/16-15469-2018-04-23.html

Neill, Sean .O, (2019), “How Creative is your Computer”?,New Scientist. https://slate.com/technology/2014/12/lovelace-test-of-artificial-intelligence-creativity-better-than-the-turing-test-of-intelligence.html.

Palm G. (1986) “Warren McCulloch and Walter Pitts: A Logical Calculus of the Ideas Immanent in Nervous Activity”,  Brain Theory, Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-70911-1_14

Raquel Acosta, Raquel. (2012),  Artificial Intelligence and Authorship  Rights,  available at http://jolt.law.harvard.edu/digest/artifical-intelligence-and-authorship-rights.

Tripathi, S., &Ghatak, C. (2018),  Artificial Intelligence and Intellectual Property Law, Christ University Law Journal 2018 Vol.7, No.1, 83-97.




Let's Start With Publication