Decoding the Report on AI Governance

Research Details

Home Research USISPF Briefer: Decoding the Report on AI Governance

USISPF Briefer: Decoding the Report on AI Governance

Background

Recognising the transformative power of artificial intelligence for a developing economy like India’s, the Government allocated a sum of almost 12 million USD for the recently launched IndiaAl Mission under the Ministry of Electronics and Information Technology. Key pillars include: IndiaAl Compute Capacity, IndiaAl Innovation Centre (IAIC), IndiaAl Datasets Platform, IndiaAl Application Development Initiative, IndiaAl FutureSkills, IndiaAl Startup Financing, and Safe & Trusted Al. This mission’s aim, as stated by the Centre, was to “ensure a structured implementation of the IndiaAl Mission through a public-private partnership model aimed at nurturing India’s Al innovation ecosystem.¹”

However, this financial outlay requires support at the policy and regulatory level. In order to ensure that the benefits of Al are equitably distributed and that any significant harms are sufficiently captured, there is a pressing need to put in place governance mechanisms. Recognising this, an Advisory Group, chaired by the Principal Scientific Advisor, was constituted on November 9, 2023, in order to undertake the development of an ‘Al for India-Specific Regulatory Framework’. The Subcommittee was headed by Dr Balaraman Ravindran, chief of the department of data science and Al and the Centre for Responsible Al at IIT Madras. The ensuing report on Al Governance, published for public consultation on January 7, 2025 aims to guide the development of a trustworthy and accountable Al ecosystem in India. The last date for providing public comment is the 27th of February, 2025.

Approach of the Subcommittee

The Subcommittee examined key governance issues, conducted a gap analysis of existing frameworks, and proposed a comprehensive approach to ensure the trustworthiness and accountability of Al systems. The Report establishes the necessity for a coordinated, whole-of-government approach to enforce compliance and ensure effective governance: the alternative is a fragmented approach to Al across ministries, which is less than advisable. It also suggests that, since there are different risks across the Al lifecycle (development, deployment, and diffusion of Al systems), people would best be served by a nuanced approach that takes that into account. It also emphasised the need for transparency and responsibility across the whole ecosystem, instead of a silo-ed approach/a sectoral view.

The Report lays out eight key principles that should guide Al governance, as follows:

The Subcommittee also performed a gap analysis, in order to identify and take stock of any gaps in current legislation (that were likely to exacerbate harms). For instance, the proliferation of deepfakes/synthetic media, which could be addressed by existing law (the Information Technology Act, the Indian Penal Code, or the Prevention of Children from Sexual Offences Act). Along with these, the existing cybersecurity legislation (such as the Digital Personal Data Protection Act), along with technological measures like watermarking, might prove sufficient to tackle these harms. Applying existing laws is preferable to instituting new ones. This is consistent with the general theme of the Report, which favours a light-touch, agile approach to regulation (because regulation carries costs).

In addition to these previously enumerated harms, the Subcommittee examined the current IPR (intellectual property rights regime) in India, to see if it would be adequate to tackle copyright harms arising from Al. In the event that Al models are trained on copyrighted data without the requisite approvals, there is a possibility that the output produced by such models would lead to infringement. The Subcommittee floated the idea that certain guardrails may need to be instituted in order to protect rights holders. Other questions raised as pertinent were the suitability of granting copyright to works generated by Al and how to tackle biases/discrimination.

Key Recommendations

The Report makes certain principal recommendations as the way forward.

These recommendations are both forward-looking and practical, and not intended to be rigid or unnecessarily punitive (even the Al incident database is intended to encourage reporting rather than to penalize the people who report). Additionally, they are envisaged as deep, long-lasting collaborations between industry and the government.

The Report mentions the necessity of technical workshops and collaborative discussions with industry (all led by relevant regulators and government departments). The Technical Secretariat could also lend cross-sectoral expertise and maturity to these efforts (since it would be staffed by “existing Meity officials as well as lateral hires, young professionals, and consultants.” This is a necessary measure, since the IT ministry currently does not have the capacity to take on such a task. Capacity-building is the need of the hour and the only way to act on the many suggestions laid out in the Report.

Subscribe to our
Daily Newsletter Weekly Newsletter