[Jishnu M Nair is a Senior Attorney at IBM]
With a score of 63.67, India is at the 32nd position in the Government AI readiness Index 2022, issued by Oxford Insights. The index symbolizes that India is more policy-ready than Brazil and Greece, and less policy-ready than China and Malaysia. Although 32 might not be a desirable rank, for a developing nation to be in the first 25% of the countries in the world in terms of ranking indicates India’s consistent efforts, especially that of the central policymaking body, NITI Aayog. The economic benefit of artificial intelligence (AI) is proven beyond doubt. It is also important to note the impact on the layperson. For the 1.3 billion people in India and especially the 845 million rural population, this would mean financial inclusion, better traffic administration, faster processing of tax forms, or better access to a health infrastructure. Technologies such as AI can resolve several of India’s challenges, and it makes sense for India to prioritize policymaking in this area.
The Indian Strategy on AI
India’s national strategy on AI and the work towards a more centralized policy are in the right direction. However, India’s lack of experience and a cautious approach have impacted its agility to implement these technologies. Banning a practical innovation is a general norm by many governments worldwide. If you don’t understand something, then you ban it. NITI Aayog deserves credit for taking a more positive step in this direction. These promising developments raise questions about what accountable and ethical AI deployment in India would look like.
Like any other technological innovation, AI is full of challenges for regulators. An instance of concern about AI is facial recognition technology (FRT), one of the most controversial of AI technologies. FRT is supposed to be used with legal authority, regulation, and oversight. The possibility of violating fundamental rights such as the right to privacy, freedom of speech and expression, and the right to protest peacefully, is high. A step in this direction was IBM’s call to ban technologies like FRT for mass surveillance, racial profiling, violations of fundamental human rights and freedoms, or any purpose which does not align with the stated principles regarding trust and transparency for AI. A case study on this is the tender released by the National Crime Records Bureau under the Indian Ministry of Home Affairs, which seeks to adopt a facial recognition system called the National Automated Facial Recognition System. The expectation was to create a surveillance tool by linking crime databases and facial recognition. There are assertions that the Home Affairs ministry structured the tender without considering privacy impacts or considering topics such as discrimination, feasibility, and exclusion caused by such facial recognition technology used by the Government, which are, by their very nature, mass surveillance measures. In India, specific laws about FRT or personal data protection do not exist.
Supreme Court’s Views
The Indian Supreme Court, in its landmark judgment of Justice KS Puttaswamy v. Union of India, laid down principles under which privacy is recognized as a fundamental right and included a need to have legislative safeguards to protect privacy, especially for the data set available to technologies. AI implemented without legislative principles or executive guidance violates the Indian Supreme Court’s decision.
On the one hand, India ought not to be in a situation where it blindly rejects new-age technologies and their implementation. It needs regulated sandboxes to provide a safe and regulated environment for the technologies to be researched and implemented. India does not have any form of framework on how government departments can structure technology procurement. There is indicative guidance for technology procurement issued by the Ministry of Information Technology, which is rarely followed while departments make their procurement.
AI is increasingly adopted in India across various sectors. The government of Andhra Pradesh has implemented a project called “Precision Farming using Drones and AI” to help farmers optimize the use of resources such as water and fertilizers. Drones with sensors collect data on soil conditions and crop growth, which is then analysed using AI algorithms to provide recommendations on optimal resource use. The Tamil Nadu Government is helping farmers to diagnose the pest infection in their crops and provide them with remedial measures. A farmer can click a picture of the pest-infected crop even with a low-cost mobile camera and upload it to the Uzhavan app. Once the photo is uploaded, the inbuilt intelligent system analyses and identifies the pest and sends the remedial measures as a text message in Tamil to the farmer’s phone.
Another effort in public works is a low-cost rural ‘drinking water supply monitoring system’ that will give a daily report on the drinking water supply in rural Tamil Nadu. It is a low-cost rural drinking water supply and monitoring system using cost-effective IoT sensors. Another example is the Indian Railways using AI to predict equipment failures and prevent accidents. AI algorithms analyse data from sensors on trains and track infrastructure to identify potential problems before they manifest, allowing maintenance to be carried out before failures occur. In another case, facial recognition is used in government offices and schools in India to track attendance and predict where attendance falls short.
Many examples indicate that it is high time we move from conceptual discussions on creating guidelines for procurement to actual implementation. India has no policy framework for its prospective engagement with AI technologies. We need a uniform public procurement framework for AI, giving directions to government departments on AI-based technologies’ procurement. Government procurement rules and purchasing practices always strongly influence markets, particularly in their early development stages. The Government can use its moral authority and credibility to set the baseline and debate AI standards. Technology providers have working models to understand the existing algorithmic challenges, and the Government can look to them for their expertise to create a baseline. These technologies’ creators look to governments to create clarity and predictability about managing the challenges, including regulatory challenges. The AI industry will lead this part of the policymaking with government agencies. Governments must proactively shape AI technologies’ development and deployment, as this conversation cannot be left to the industry alone. Therefore, we need common-sense frameworks to help governments overcome reluctance and mistakes in procuring complex new technologies and tackle a new market. Transparency in guidelines would mean a level playing field for permitting established companies and new entrants to the AI space. India can consider emulating the existing World Economic Forum’s (WEF) AI Government Procurement Guidelines for its AI procurement activities in this context.
What are the WEF AI Government Procurement Guidelines?
The WEF AI government procurement guidelines (“WEF Guidelines” or “Guidelines”) set forth procurement processes that concentrate on outlining problems and opportunities and thereby leave room for iteration. The Guidelines have sought to address the concerns around AI, such as bias, privacy, accountability, transparency, and overall complexity. Amongst other things, the WEF Guidelines insist on creative procurement processes, including considering challenge-based procurement and establishing a well-constructed and detailed framework agreement that the suppliers can join. This would be a game-changer in a country like India, where we have multiple forms of government tenders. The other expectation is for the procuring organization (“PO”) to do an initial AI risk and impact assessment.
The Guidelines also suggest that AI-powered solutions should stand on the pillar of transparency. In a procurement process, POs should expect the AI supplier to submit documentation regarding the development of the algorithm. The document should include headers like the data set used for development, whether the model is based on supervised, unsupervised, or reinforcement learning, or any known biases. If an algorithm is building a process, especially an administrative process, it should be established with the supplier regarding how the general public can contest automated decision-making. We need to ensure that AI decision-making is as transparent as possible, explore mechanisms to enable interoperability of the algorithms internally and externally and provide no vendor lock-in.
The Guidelines also expect clearly defined data governance mechanisms in place from the start of the procurement process. They ask that the procurement process consider the data’s susceptibility in scope and if its usage is fair. The vendors are expected to highlight known limitations (e.g., quality) of the tender data and require tenderers to describe their strategies to address these shortcomings. The guidelines also hope the PO makes ethical considerations part of their procurement process, especially evaluating the technicality around prospective suppliers’ proposals.
The WEF Guidelines provide fundamental considerations a government should address before acquiring and deploying AI solutions and services. The Guidelines suggest structures that align with ethical principles and human rights. They explain steps to evaluate the potential risks and impacts of the AI system and put in place measures to mitigate any negative consequences. The Guidelines can also ensure that the procurement process is transparent and open and that relevant stakeholders are involved in the decision-making process, which in turn can ensure that the systems can be made explainable and accountable, with precise mechanisms in place for transparency and oversight.
The Guidelines themselves say they are not intended as a silver bullet for solving all public sector AI adoption challenges. Still, by influencing how new AI solutions are procured, they can set government use and adoption of AI on a better path. Although the WEF Guidelines might not work in their entirety in India, the critical pillars defined above can be the guiding force in building up an effective regime for the Indian government’s foray into AI procurement.
– Jishnu M Nair