Need for regulatory framework in today’s AI-driven data ecosystems

The recent global Partnership for Artificial Intelligence, GPAI, Summit 2024, held in Belgrade, Serbia, was witness to growing concerns around AI safety and related regulations. With Europe’s predilection for regulations that often tend to privilege protection against harms over freedom to innovate, the EU’s AI Act consultations and the upcoming AI Action Summit in Paris dominated the conversations.

Regulatory thinking in India, it would seem, stands somewhere in between the EU’s approach and the approach of the US, which has traditionally privileged innovation over overly burdensome regulations. The India AI Mission has gained momentum with the Prime Minister Narendra Modi-led NDA government in its third term pushing ahead with efforts to create a public GPU cloud and the mission to make public datasets. However, a debate rages on where India must place its priorities between creating India-specific sovereign foundational models or whether India must build on its manpower strengths as the use case and AI applications capital of the world while leaving the core technology development to big-tech majors elsewhere. It is in this context that one must address the question of what kind of regulatory framework India needs to put in place at this time.

While AI’s potential to revolutionise industries is paralleled by risks that, if left unchecked, could jeopardise societal well-being, we in India have to be seized of issues such as algorithmic bias and data misuse given the manner in which the current foundational models have been trained, glossing over ethical dilemmas of copyright issues as well as concerns over cultural and linguistic diversity. In human resource-rich India, an additional area of concern is AI’s disruptive influence on employment, given the likelihood of automation replacing certain types of jobs. The challenge for the Indian government would be to evolve the regulatory framework along pathways that do not stifle innovation but channel it responsibly without compromising public and national interest.

AI is unlike all previous waves of technology, given its propensity for not just non-linear but potentially exponential growth. While historically in India regulations have always lagged behind technological advancement with regulators having to play catch-up, in the case of AI, the capacity and skills gap is both steep and wide. Much like the manner in which global tech platforms for social media and video sharing have created a borderless world for content dissemination and foreign influence, the open availability over the internet of AI models has meant that the digitally savvy Indian consumer is several generations ahead of Indian regulators in the use of technology.

Between the amendments to the IT rules and the Digital Personal Data Protection (DPDP) Act, the NDA government has laid a regulatory foundation for managing the AI-driven data ecosystems. However, this foundation will be stress tested on several fronts and will need to evolve with agility and technical sophistication with the use of regulatory sandboxes to keep up with the trajectory AI systems will likely take. It is important to recognise the following India-specific imperatives that must guide the regulatory priorities within India, unlike the developed West, where AI safety and fears of an all-powerful AI have dominated the regulatory debate.

Firstly, it is important that we in India recognise that for AI models to adequately address uniquely Indian needs, they will require a vast body of publicly accessible Indian datasets. From written text to spoken words, from visual imagery to cultural artefacts, we will not only have to digitise India’s diverse socio-cultural contexts but will have to put in place a framework by which Indian public bodies, institutions, corporations, and organisations are motivated to make their datasets public voluntarily. A call to action for "Data Daan" is the need of the hour to build a national movement on donating data for a digital India, where AI models accurately reflect the Indian context. Such a national call to action could also empower Indian developers and researchers to address the requirements of uniquely Indian use cases through indigenous models.

Secondly, it would be to India’s advantage to tap its vast human resource pool to provide value-added data services focussed on annotating and enriching these public datasets to aid and accelerate in model training and development. Data enrichment services that require both data skills and specialised knowledge of specific domains could drive the next wave of AI-era jobs but could also position India as a global service provider for these services requiring deep domain knowledge.

Lastly, vast portions of India’s public sector that have lagged in digitalisation have the unique opportunity to leapfrog into the AI era without the baggage of legacy IT systems and unfriendly interfaces, which have largely contributed to the digital divide. From education to local governance, we have the opportunity with AI to bridge the digital gap in the last mile of public service delivery by overcoming language barriers and capacity limitations of our public bodies through a new class of AI-enabled digital public goods.

India, with its demographic dividend and technological prowess, can strike a different regulatory path that not only embraces but also safeguards societal interests. It has the opportunity to show the world the way to keep developmental priorities front and centre in the evolution of AI systems.

Views are personal. The author is Shashi Shekhar Vempati, Co-founder, AI4India.org; Former CEO, Prasar Bharati