How India is shaping up its first generation of AI laws

From creating a datasets platform under India AI strategy to signing the Bletchley Declaration, regulations are the centrepiece of debates.

India’s tryst with regulating Artificial Intelligence (AI) has only begun and is expected to grow further as it proliferates every aspect of life. In this process, however, AI has also shown how it can be leveraged for the bad, such as in deepfakes strewn across social media. All of this has prompted the Government of India to focus on regulating AI as a standalone entity, in a bid to channel its development in the right avenues and penalise misuse.

Speaking at the recent Confederation of Indian Industry (CII)’s Global Economic Policy Forum, S Krishnan, Secretary of the Ministry of Electronics and Information Technology (MeitY), highlighted that from the policy perspective, India believes that AI offers “more positives than negatives” and must therefore be leveraged for constructive tasks.

Highlighting that AI has the potential to have a transformative industry impact the way the Industrial Revolution did, Krishnan said that it is this potential that India wants to tap into. To do this, the Centre has identified four key challenges: access to compute, building organised data sets, developing skills, and prompting nationwide and global collaborative research on AI. Of these four, while India is at par or ahead of global counterparts in data availability, skills and research, it is access to compute across enterprises that presently poses a challenge.

Bias, too, presents an issue that India’s own AI journey will look to tackle. To do so, India, during the Global Partnership on AI (GPAI) Summit in New Delhi between 12-14 December, will pitch to attract more nations to become a member of this global collaborative effort to regulate AI. “The membership is largely driven at present by OECD nations. India is looking to get more of the global south involved here, to address the unique challenges that will need to be factored into AI development in the long run. This is crucial,” Krishnan said.

“If we can train models in that way will be more suited to the rest of the global south as well, that is another important dimension. Our efforts towards Bhashini and a variety of others are in this regard,” he further added.

To be sure, this approach is in line with how the Centre seeks to draw the line when it comes to AI development. Historically, MeitY spokespeople have reiterated that India will not look to strictly regulate AI. Instead, it will only look at regulating the harmful aspects of AI. Krishnan, too, reiterated this approach.

“India is looking to get more of the global south involved to address the unique challenges that will need to be factored into AI development in the long run.”- S Krishnan,Secretary, Ministry of Electronics and Information Technology. Government of India

Bletchley Declaration seeks to define a common framework for building AI so that the technology does not get misplaced or is developed without guardrails.

The upcoming Digital India Act (DIA), which has now been stalled until the formation of the next government, is expected to handle how AI harm is regulated. The Digital Personal Data Privacy (DPDP) Act, which has already been notified, already announced guardrails for data collection in the development of AI.

On 13 October, Rajeev Chandrasekhar, the Union Minister of State for Electronics and Information Technology, said that a new report on the India AI strategy will look to define a roadmap for the creation of organised, centralised datasets for multiple Indian languages, and also establish a plan to build indigenous compute power for AI tasks. The datasets project will involve lending access to Indian language data parameters to researchers, which will be crucial since the availability of data in non-English languages is still not elaborate.

Such datasets, however, are building—platforms such as OpenAI’s Generative Pre-trained Transformer (GPT) models already support Indic languages, while Google’s previous AI model, PaLM, also supported Hindi and other languages.

India, on this note, also became one of the first 28 signatories of the first global alliance on regulating AI. Called the Bletchley Declaration and signed in the UK on 1 November, it seeks to have nations, including the US and China, agree on a common framework for building AI so that the technology does not get misplaced or is developed without guardrails.

Speaking to the media after the UK AI Summit last month, Chandrasekhar shared that the Indian delegation stressed the need to have safe, trusted AI platforms, and distinguish them from unsafe, untrusted platforms. “We proposed that AI should not be demonised as it represents a massive opportunity, in India, as around the world. We also spoke about who will determine safety and trust and discussed four harms that come out of AI. These include workforce disruption, privacy impact on individuals, non-criminal harms, and finally, weaponisation and criminalisation of AI.”

Summing it all up, Krishnan said that India’s AI regulatory approach will be “light touch”, akin to how it enabled the domestic IT services sector to flourish in the 1990s. By preventing harm, India seeks to enable the industry to pursue innovation in the field, while the development of local computing power could bolster these initiatives further.

By Vernika Awal

feedbackvnd@cybermedia.co.in