INTRODUCTION
“The real question is, when will we draft an artificial intelligence bill of rights? What will that consist of? And who will get to decide that?”
• Gray Scott, Futurist & Technologist
Machine learning algorithms, predictive analysis, automated decision-making systems and large language models are few of the components that form a part of the wide range of AI technologies, that aren’t just limited to research labs anymore, but have spread widely across an array of fields in India including healthcare, finance, maintenance of law and order, governance, education and even agriculture. The Central Government of India as well as multiple State Governments have promoted the campaign “AI for all “and have been in favour of digital transformation. However this digitalization and broad application of AI across innumerable sectors of life, come with their own sets of risks and dangers such as – algorithmic bias, automated decisions that directly impact rights, data privacy breaches, replacement of human beings in the work field, surveillance and accountability gaps, which must also be taken into consideration in order to form a dedicated and overreaching legal system in India with respect to artificial intelligence and technology, which is the need of the hour. The absence of a substantial legal framework considering AI, poses a risk to not only individual rights but also places a question mark on transparency, trust and legitimacy of the AI systems that are in use on the private as well as public front. The Constitution of India guarantees the rights to life, liberty and equality to every individual irrespective of age, sex, race, caste or creed and assures protection from social inequality. Given this context, strict rules, regulations and a proper legal framework in place have become a topic of not just academic discussion but an urgent concern demanding immediate attention. The researcher discusses the need for India to urgently adopt a strict legal statute concerning AI technologies.
ISSUES RELATING TO AI IN INDIA
The widespread uses and applications of AI technologies in a wide variety of sectors as well as day-to-day life in the Indian scenario, brings with it the following issues and challenges:
1. India is a land full of diversity in various aspects be it languages, cultures, religions, castes and many more. AI technologies have found their role in processes such as credit scoring, recruitment and admission procedures, etc. These technologies pose a risk of aggravating rather than alleviating socio-economic inequalities based on gender, caste or region.
2. Plenty of AI systems, specifically deep-learning models are termed as “black boxes”. AI systems lack transparency and cannot be held accountable, if need be, hence, when such AI technologies are required to take decisions as in cases of social welfare, medical diagnostics, parole or law enforcement, the outcomes may hinder minutely or severely with people’s rights owing to such lack of transparency and accountability.
3. Artificial intelligence systems and technologies feed on large amounts of data that may often be personal or sensitive in nature. India has recently enacted the Digital Personal Data Protection Act, 2023, to safeguard and prevent the misuse of personal or private data. However, with the advent and wide range of application of AI technologies, a whole new set of challenges including those of consent, misuse and theft of identity, profiling, reuse of data and its interpretation, are peeking up.
4. AI systems and technologies have found their way into major decision-making systems. Any misjudgement on its part such as – wrong medical diagnosis, misinterpretation leading to wrong decisions thus causing accidents by autonomous vehicles and drones, or wrongful denial of benefits, leads to the question of responsibility and liability. Who would be responsible for the aforementioned harm and damages – the developer, algorithm or deploying agency? Indian laws have no clear mention of the same.
5. AI technologies are widely used for processes such as facial recognition, automated decision making on a large scale, predictive policing, etc. The Constitution of India guarantees the rights to equality, life and liberty to all, irrespective of gender, caste, creed, etc. under Articles 14 and 21. The use of AI technologies in the above-mentioned scenarios, threatens the above rights as well as the right to privacy, as has been held by the Court in Puttaswamy. However, the Indian legal system currently lacks specific measures for safeguarding and protecting these rights.
6. India is devoid of a horizontal regulatory sandbox catering to AI technologies and also needs well defined ethical standards or certification system for the regulation of AI applications, like Fintech and other regulated sectors.
7. A lot of AI technologies in India deploy cloud-based data systems and algorithms that may have cross-border implications, leading to challenges and issues pertaining to governance, jurisdiction and enforcement, owing to a lack of clearly demarcated laws.
INDIAN LAWS DEALING WITH AI
India is devoid of specific laws catering to AI and related issues; however, various regulations and legal frameworks deal with the said issues indirectly:
• IT (Information Technology) Act, 2000:
The rules and regulations mentioned in this Act, such as the Intermediary Guidelines, form the base level digital legal framework and include areas such as e-commerce, cybercrimes, data offences.
• Digital Personal Data Protection Act, 2023 (DPDP)
This Act serves to be a foundation for data regulatory legal frameworks; while not specifically dealing with AI, it plays a pivotal role in data regulation by AI technologies. This Act deals with rights pertaining to data principals, digital processing of personal data and obligates agents dealing with data.
• Sectoral regulations
The SEBI (Securities and Exchange Board of India) dealing with the financial sector, has released consultation documents regarding the use of AI/ML in the securities and exchange market.
• Policy and guidelines
The NITI Aayog proposed the “National Strategy for Artificial Intelligence”, in 2018 and followed it up with “Responsible AI for All”, in 2021. These are a few examples, that though aren’t binding laws themselves, form a basic architectural framework for framing regulatory policies and guidelines and forming legislations pertaining to AI technologies.
• The Consumer Protection Act, 2019, from the purview of AI, is applicable if faulty AI algorithms lead to wrongful decision making, thereby negatively impacting consumers.
• Intellectual Property Laws-
Intellectual Property laws do not directly and specifically deal with AI systems and technologies. However, they become relevant in the events when AI technological training data makes use of content that has been copyrighted.
• The Telecommunications Act, 2023
This Act deals with AI based incidents in the telecom industry. It proposed regulatory legislations for the same, however, these legislations do not specifically cater to AI systems and rather act as a partial regulatory framework.
GAPS IN THE ABOVE EXISTING LAWS DEALING WITH AI
The above-mentioned laws deal with AI systems and technologies indirectly and thus, several significant lacunae exist, which are as follows:
India is lacking in an intricate and specifically dedicated legal framework, dealing directly with the governance and regulation of AI technologies. The legal statute must take into account the following points – definition of Artificial Intelligence, classification of the risks of AI systems, assessing impacts of AI technologies mandatorily and assessment of certification, audit and transparency obligations.AI systems and technologies have become advanced enough to take decisions autonomously, though some decisions may not be accurate and appropriate and may lead to more harm rather than the good it intended to do. There isn’t any law that clearly and specifically allocates responsibility, whether to the developer, of deployed, of vendor, or user, for any harm that might arise owing to the misjudgement on the part of AI technologies.
Also, there is a sheer lack of any law, which makes it mandatory for AI systems with high risk to be explainable, accountable or include assessment systems for studying algorithmic impacts. The Constitution of India guarantees equality to all irrespective of gender, caste or creed. However, AI technologies and systems deployed for recruitment of jobs and academics, regulation of welfare, credit systems and law enforcement may sometimes threaten this Constitutional right and there is no specific legislation to safeguard the same.AI technologies are still budding, meaning they’re in a phase of intense experimentation and innovation. India lacks a dedicated and intricate legal framework to tackle challenges that may arise as a result of these technologies. There is a lack of regulatory bodies, certification procedures and measures to keep these systems and technologies in check, ethically or otherwise and thus, there exists a legal grey zone, leaving scope for AI developers and innovators to establish loop holes.
Furthermore, AI technologies have found their role in the governance and welfare processes such as predictive policing, welfare eligibility, etc. However, the State’s use of these technologies does not have a counter-regulatory legal framework in check, in case questions relating to transparency, oversight, redressal, audit of independent review arises. AI systems and technologies have found a wide array of applications in India in a variety of fields including health, finance, agriculture, governance and so on. Multiple issues and problems also crop up with such a vast application of these technologies. However, India lacks an intricate framework to regulate and coordinate this cross linkage of AI technologies across various sectors.
Specially as the world has become extremely fast-paced, AI technologies and systems are also evolving at a rapid speed. However, the current laws and legal frameworks are termed as ‘technology-agonistic’, as they tend to be rigid and fail to keep up with rapidly evolving pace of artificial intelligence technologies. The outreach and effective application and enforcement of Indian laws in the global AI systems and technologies scenario is still a question mark.
SUITABILITY OF ADOPTION OF MODEL AI LAWS AND REGULATION BY INDIA
The mix of following models can be considered in order for India to form an AI-centric legal statute
European Union model
The European Union has put forth the idea of an AI Act that brings forward a risk-based classification of AI systems, which includes the categories of risk as Minimal risk, Limited risk, High risk and Unacceptable risk.
The Act promotes and allows innovation while keeping a check on human oversight, making transparency with respect to AI technologies an obligation, assessments and checks that are mandatory to ensure ethical evolution of AI, thereby laying emphasis on protection and safeguarding of fundamental rights.
United States model
The United States of America focuses on aspects such as guidance, voluntary standards and a harmonious collaboration between private and public sectors, thus taking a collaborative sectoral route, involving agencies like the FTC, SEC, FDA, etc. Thus, the US model focuses on bridging and coordinating multiple factors, rather than harbouring on a single law. As a result, innovation of AI technologies becomes a rapid process, but protection of rights may be compromised.
China model
China focuses on achieving control, proper surveillance, governance in alignment with State goals and strict control and regulations over certain malpractices associated with the use of AI such as deep-fakes, social credit, etc. Thus, China has an overall State led approach for governance.
Recommendation for India
After studying all the three existing models with respect to legal frameworks and regulations for AI technologies, it is opined that India should adopt a hybrid model, forming an amalgamation of all the merits of these models, viz.
• The European Union’s horizontal legal framework, having a risk-based approach taking into account transparency, audit, protection of rights, imposing moral and ethical obligations as imperative for the use of high-risk AI technologies.
• The flexible and innovation promoting approach adopted by the US, which allows and promotes experimentation and evolution of these technologies
India being a diverse country, the regulatory model must cater to the current socio-economic scenario and address topics such as resource constraints, plural society, constitutional rights, digital divide and must focus on:
• An Act dealing solely and specifically with AI systems and technologies, taking into account the definition of AI systems, risk stratification, making impact assessment of these technologies mandatory and obligatory audits for high-risk AI applications such as in fields of healthcare, crimes and justice, welfare eligibility, autonomous vehicles, etc.
• Setting up a regulatory sandbox or establishing a certification process to regulate and keep a check on AI testing and innovation.
• Establishing an independent authoritative body to deal with AI technologies, having its own technical capacity to oversee, supervise, audit and coordinate among various sectors.
• Defining specific guidelines for the various sectors witnessing the use and applications of AI technologies such as, finance, healthcare and law enforcement.
• Enhancing provisions dealing with liability, transparency, accountability and fairness concerning the algorithms on which the AI systems and technology’s function.
• Upholding constitutional rights, specifically Articles 14,19,21 and the data protection regime (DPDP Act)
• Providing incentives, thus promoting innovation while maintaining adequate enforcements and checkpoints to maintain accountability.
Thus, a hybrid approach is recommended to ensure the promotion of innovation while protecting rights, keeping in mind India’s developmental goals and priorities and also its governance capacity.
ROLE OF THE JUDICIARY WITH RESPECT TO AI IN INDIA
The Indian judicial system, particularly the High Court and Supreme Court are pivotal in regulating the AI ecosystem:
1. In the case of K. S. Puttaswamy V. Union of India, the Indian judiciary has already recognised the right to privacy. The implication of these rights, namely- Article 14 (Right to Equality), Article 19 (Right to Freedom), Article 21 (Right to Live and Liberty), with respect to AI technologies needs to be studied and applied in scenarios involving AI based decision making.
2. With respect to the scenario of application of AI technologies in sectors such as facial recognition, policing, welfare allocation by the government, the Courts may play an instrumental role in keeping the accountability, fairness, lawfulness and transparency of such systems in check and ensuring that adequate safeguards exist.
3. Taking into account the current legal scenario, where there is a lack of adequate legislation, the Courts may come up with landmark judgements dealing with AI based liabilities, responsibility of the developer/agency, negligence occurring in framing the algorithms, product liability analogies.
4. The Courts can demand justifications for automated decisions taken by AI systems and technologies, make audits mandatory and even come up with remedies and redressals dealing with discrimination or unjust decisions caused by AI algorithms.
5. By coming up with breakthrough landmark judgements, the Courts may direct regulators and amend the legislations to form SOPs and bridge the gaps in regulations dealing with the use of AI technologies.
6. Even after regulations and legislations concerning AI systems and technologies are in place, the Courts shall play a pivotal role in their interpretation, adequate enforcement and resolution of any ambiguities.
Thus, the Indian judicial system plays a pivotal role to make sure that the regulatory legal team framework remains not just on paper but also takes shape in reality in terms of jurisdiction of AI systems and technologies. With respect to the Indian scenario, where judicial activism is instrumental in the protection of rights, the role of the judiciary is indispensable.
CONCLUSION & SUGGESTIONS
The advent of AI systems and technologies and its widespread application in a huge amount of sectors in India, brings with it not just a humungous promise but also poses significant risks. These technologies boast of significant hope to rampantly speed up innovation and development, improve efficient delivery and utilisation of public services and open up a wide array of golden opportunities in various fields and walks of life. On the flip side, the Indian legal and regulatory statute is still divided and lacks an AI specific nature, missing out on essential aspects such as transparency, liability, accountability and bias mitigation. These lacunae pose a significant risk to the rights of the population, thereby threatening to weaken trust in these technologies and thus hinder in their large-scale adoption and application. India is currently faced with only two options – to either frame a regulatory legal framework considering the importance of innovation and the future, balancing it with protection of the guaranteed rights or to continue with adhoc laws and regulations, hoping for them to be effective and sufficient. An urgent decision is the need of the hour as AI systems and technologies are spreading their wings and finding their way into a large number of sectors at rapid speed. Coming up with legislations dealing with AI, having sectoral flexibility in synergy with a strong, efficient regulatory authority, mandatory transparency and audit obligations, a regulatory sandbox, thus adopting a hybrid approach looks promising. The judicial system as it has all along, shall continue to be instrumental in the effective enforcement, regulation and interpretation of this regulatory framework. India still lacks such a broad looking and comprehensive approach towards AI systems and technologies and effective regulations and legislations for the same, thereby posing a great risk upon its citizens as well as the innovation ecosystem, leaving them exposed and ungoverned.
Suggestions
The researcher would like to make the following suggestions after careful study and analysis as mentioned above:
1. Sets definition for terms like “deployment” “high risk AI” “algorithmic decision making” and so on.
2. Provides risk-based stratification of AI systems and technologies as high, moderate, low and set mandatory obligations accordingly to deal with the same.
3. Calls for impact assessments in mandatory fashion, be it with respect to algorithms, equity or human rights, before AI technologies with high risk are given a green signal for deployment.
4. Makes it compulsory for AI based decision making thereby affecting basic rights, such as a decision negatively impacting a citizen’s rights, to be explainable as well as transparent.
5. Appropriately demarcates logging, certification, audit and conformity assessment requirements.
6. Specifies accountability and liability – clarifying responsibility in situations where AI causes harm – whether upon the developer, vendor, deployer or user.
7. Makes provisions for a regulatory sandbox, so as to permit experimentation and innovation under supervision.
8. Sets up an AI regulator/authority as an independent body with multiple stakeholders to supervise enforcement, audit, implementation, standardization and coordination among sectors.
9. Establishes redressal seeking mechanisms with respect to the negative impacts of AI based decisions such as the right to appeal, explain, review.
Other Legislation in India
In coordination with the Digital Data Protection Act, 2023, the IT Act reforms as well as sectoral laws, the AI Act must also work in synergy to ensure that privacy, data protection, profiling and algorithmic decision making are well reasoned.
1. In scenarios involving high risk AI technologies such as diagnostics and healthcare, welfare automation, autonomy of vehicles, criminal justice, etc. , the legislations and regulations must be stricter, more structured and well defined while the low risk systems may be shown some flexibility in terms of regulation.
2. As a means to effectively promote and encourage AI innovation and advancements in India, while ensuring responsibility, incentives must be provided, public private relationships must be promoted, regulatory sandboxes should be set up and a standardization and certification mechanism should be put in place.
3. In order to ensure that AI systems and technologies remain fair and all inclusive and to safeguard against discrimination, algorithms should have regular fairness checks, bias audits must be conducted, outcomes amongst various socio-economic groups must be reviewed periodically and mandates should be put in place to overcome human oversight.
4. Strict guidelines must be published by the regulator to tackle AI related harms. Additionally, compliance reports must be submitted periodically, records of high risk AI deployments must be maintained and regular audits of such systems should be conducted, thus establishing a ‘whistle blowing’ system to deal with AI associated harms.
5. The judiciary plays a pivotal role in the amendment and effective application and regulation of AI related laws and thus, judges/tribunals should be trained with respect to AI systems, algorithm-based transparency, digital evidence and appropriate review standards in cases that deal with AI.
6. Stakeholders of AI systems and technologies such as industries, academic industry, civil society and marginalized communities must be consulted and made part of the legislative process. The major AI systems in State use must be entirely transparent to the general public in terms of its design, algorithm, audit results and redressal mechanisms.
7. India should attempt to contribute to global governance systems concerning AI technologies by being a part of multilateral forums. To achieve this, it should articulate with global principles such as the Organisation of Economic Cooperation and Development.
8. AI systems and technologies are evolving at rampant speed. To ensure relevant regulation of these systems, measures such as periodic review, sunset clauses and mechanisms, evaluation of emerging risks and mandatory audits and transparency should be included to enable adapting to new forms of AI such as generative AI, autonomous agents, large language models, etc.