Tech for governance

India could become a leader in use of AI

The Union minister of state for electronics and information technology, Rajeev Chandrasekhar, at an event organised by this newspaper last week, highlighted how the government was using technology in governance. Technology has helped plug leakages in subsidy programmes, and artificial intelligence (AI) could materially improve efficiency in e-governance. The Union Budget had announced the establishment of three “centres of excellence” for AI. These are to be connected with academia, industry, and start-ups. Working groups have already been set up for designing India dataset platforms, and the centres of excellence. The government is considering a hub-and-spoke network model, which will institute safeguards and “guardrails for ethical use without disrupting innovation”.

The government will spend around Rs 1,635 crore to develop this AI ecosystem to make e-governance platforms more intelligent. The priorities for deployment include governance applications of the India Stack, powering up the large language model for Digital India Bhashini, and building smarter health care services. The re-launched Skill India programme will also have an AI focus, since this will be essential for future workforces. In analogy to the US and China, where governments and government agencies funded early forays into AI, the policy thrust could spark the private sector into developing new use-cases and trigger new investments into AI-related start-ups. This secondary effect would be vital in generating higher productivity across the economy.

The India Stack is a large basket of open-source software application programming interfaces (APIs) of government-backed services such as Aadhaar, United Payments Interface, eSign, and DigiLocker. The open-source nature allows anybody to connect. This has inspired many different apps with varying architectures, APIs, libraries, and user interfaces. The Stack generates vast data across its myriad use cases. If AI-driven algorithms are applied to study that data, it could provide granular, deep insights into consumer behaviour and consumption patterns. While such AI induction may be laudable in intent, tight oversight will be required to maintain privacy and avoid data leakage, especially in the absence of a data-protection law. The precautions may include ensuring only Indian start-ups and researchers have direct access to the datasets. Beyond that, it requires strong anonymisation protocols since there is no guarantee against misuse by local organisations.

Moreover, if India uses AI at scale to power digital inclusion and skilling, it will need to develop robust filters to root out algorithmic biases, and to ensure AI doesn’t perpetuate existing biases against castes and communities present in the data. This involves creating audit systems for understanding how AI “thinks”, since it can be a black box even to the programmers. One reason why AI deployment in India datasets could be very powerful is the sheer size and diversity of the country and, therefore, of the datasets. When it comes to large language models (ChatGPT, for example), India again has an inbuilt advantage in that there is data from a plethora of languages to work with. It could become a world leader in AI if it pushes ahead with this policy. Side by side, the government could consider developing guidelines for protecting and mitigating potential harm to citizens.