The Aakhya Weekly #85 | Ctrl+Alt+Regulate: Navigating the Frontiers of AI Governance

In Focus: AI Governance: Balancing Progress with Prudence

By Sakshi Satija

Artificial Intelligence (AI) stands as a beacon of immense potential, offering solutions across diverse sectors from healthcare to finance. Yet, the complexity of AI's capabilities necessitates robust governance to address potential risks and ensure ethical deployment. Against this backdrop, the announcement by Minister of State for Electronics and Information Technology, Rajeev Chandrasekhar, on creating a global draft regulatory framework for AI, signifies a pivotal step towards striking a balance between advancement and accountability. Slated for release by June-July, the draft framework reflects a proactive approach to navigating AI governance, addressing concerns such as bias and misuse during model training.

The global nature of AI technologies necessitates a refined approach to AI regulation, highlighting the demand for a cohesive governance framework that transcends geographical boundaries. In line with this, the upcoming draft AI regulation framework could help lay the groundwork for a structured approach to AI governance in India. As we delve deeper, it is essential to define the scope and challenges of AI in today's context.

Defining AI

As AI has innumerable applications, the wide variety of tools creates hurdles in formulating an all-encompassing definition for varied systems and technologies that shall be subjected to AI regulations.

Countries that have attempted one have often fallen short – either due to their ambiguity, or narrow definitions for diverse AI tools. For instance, the US Future of AI Act defines AI as a system that works “without significant oversight” leaving ambiguity regarding the degree of human involvement that will be considered “significant”. Meanwhile, the UK National Security and Investment Act defines AI as a system that “makes recommendations, predictions, or decisions” as per human objectives. This may lead to the erroneous exclusion of AI-backed image recognition, which may not be required to generate any distinct output. At present, nearly all definitions end up omitting some of the AI systems from their scope or incorrectly include non-AI systems.

Regulating AI: Sectoral or principles-based approach

Over-governing AI and applying uniform regulations without considering the specific characteristics and applications, can stifle innovation and technological advancement. For instance, subjecting an AI-based email spam filter to the same stringent regulations as autonomous vehicles could lead to unnecessary bureaucratic hurdles. It could also impede the development and deployment of useful tools while diverting resources away from areas that require oversight. By adopting a differential regulation approach, tailored to specific risks in each AI application, we can strike a balance between promoting innovation and ensuring safety and accountability. 

A sectoral approach addresses the unique struggles of different industries like healthcare, finance, and transportation. However, it creates a fragmented regulatory landscape, as it undercuts uniformity and coordination between sectors. Varying regulatory requirements could confuse businesses operating across sectors, and pose monitoring challenges for enforcement authorities. Moreover, industry-specific regulations could struggle to keep pace with the rapid evolution of AI. For instance, the current AI regulations governing facial recognition might effectively address concerns regarding the storage of images, while overlooking data protection. It could be misused by stakeholders with vested interests who may analyze them to determine emotions without human consent, leading to privacy breaches, and further posing ethical concerns about its use.

Conversely, a principles-based approach centered on adherence to fairness, transparency, and accountability, could lead to confusion. For instance, regulators may establish the principle of fairness without specific criteria or metrics for assessing fairness. If AI systems are used in hiring processes, companies could claim to be unbiased, however, without specific guidelines on what constitutes fairness, ambiguity could impact the assessment process. Additionally, regulatory enforcement could be manipulated to their advantage.  Companies could also provide selective or incomplete information about their AI systems, posing challenges to regulators in identifying risks or violations.

For the global framework, the Indian government is planning to adopt a principles-based approach over a sectoral one. Beyond this, the government also hopes to address critical issues like bias and misuse during model training, in case there are ambiguities in the principles of the proposed framework.

Accountability and data bias 

Once clear metrics are established to evaluate principles and guidelines for AI, the government also needs to decide how to determine violators of these principles and guidelines. One may be inclined to blame AI developers, however, it is crucial to recognize that the training data often contains inherent biases that violate fairness principles. Imagine a digital lending platform in India that uses AI algorithms to determine loan eligibility. If the training data for this AI model predominantly includes applications from urban areas, it may erroneously conclude that loan recipients must come from cities, while systematically rejecting loan applications from villages. This bias could result in financial exclusion for rural populations and perpetuate economic disparities, due to no intention of the developer. 

Furthermore, assessing compliance can be challenging as some AI systems operate in “black boxes”. It means that despite our knowledge of the inputs and the subsequent results, we lack visibility on how AI arrives at those outcomes. Without understanding how an AI system makes decisions, it is difficult to determine whether those decisions are fair and unbiased. Biases in the algorithms or data used to train the AI could lead to discriminatory outcomes, making it a complex challenge.

Charting a path forward

At pace with the evolution of AI, additional risks, challenges, and opportunities are unfolding. To address these challenges, collaboration among governments, industry leaders, and other stakeholders is essential to craft regulatory frameworks to promote innovation, while maintaining safeguards. In this respect, periodic reviews may enable regulators to adapt to new-age solutions, identify emerging issues, and promptly address unforeseen consequences like data or algorithmic bias. 

Additionally, ensuring affordability is important to foster widespread access and equitable distribution of AI’s benefits across society. Government intervention plays a vital role, as incentives in R&D, competition promotion, and policy implementation, could improve access and bring down costs. Furthermore, governments can support AI training and educational initiatives for free or at affordable rates. By prioritizing affordability, governments can ensure universal access, democratize AI’s benefits, and foster inclusive economic growth.

Top Stories for the Week

Maharashtra's Third Attempt to Advance Maratha Quota

On Tuesday, the Maharashtra government unanimously passed the Maharashtra State Socially and Educationally Backward Bill 2024 which mandates a 10% reservation for the Maratha community in educational institutions and government jobs. This decision, made during a one-day special Vidhan Sabha session, was based on a report from the Justice (Retd.) Sunil B Shukre-led Maharashtra State Backward Class Commission (MSBCC). In defence of the legislation, the commission referenced similar acts in Bihar and Tamil Nadu to present their case.

The urgency to pass the bill was primarily sparked by the hunger strike led by the Maratha quota activist, Manoj Jarange, which commenced on February 10th. Upon the passing of the bill, however, Mr. Jarange pointed out that the percentage of reservation mattered less than the category it is under. He advocated for inclusion within the Other Backward Classes (OBC) category, rather than separate reservation. 

In Maharashtra, 10%-reservation is already allocated for the Economically Weaker Section (EWS), predominantly benefiting Marathas who claim up to 85% of this reservation. The Maratha community's reservation efforts have previously faced setbacks in 2014 and 2019. A common thread is that all three bids, including the recent one for reservation, arrived in the run-up to elections. This is seen by several analysts as a political attempt at reaping electoral benefits from the long-standing issue. In 2021, the Supreme Court ruled against it, citing the breach of the 50% quota limit. 

The Maratha Reservation Bill will now be tabled in the Legislative Council for assent, after which it will become law. However, its legality, like in the case of the previous attempts, may face challenges from the judiciary, making the journey ahead uncertain.

With the Greek Head of State’s historic visit, Indo-Greek ties are set to take off 

On the 20th of February, Kyriakos Mitsotakis, the Prime Minister of the Hellenic Republic (Greece) arrived in India as part of his state visit, where he also inaugurated the Raisina Dialogue with a keynote address. The Raisina Dialogue is a flagship foreign policy initiative of the Indian think tank ORF in partnership with the Ministry of External Affairs. PM Mitostakis’ visit makes him the first Greek Head of State in 15 years to make a state visit to India. In August 2023, PM Modi visited Athens for a State visit where the states’ mutual priorities on defense were discussed, showing that the Indo-Greek relations are witnessing unprecedented growth.

The Greek Prime Minister was feted by PM Modi with a ceremonial guard of honor at the Rashtrapati Bhavan. PM Mitsotakis met with Prime Minister Narendra Modi to discuss defense, and mutual priorities in trade and migration policies. The two countries also concluded agreements on doubling bilateral trade to $ 4 billion by 2030 and increasing linkages between the nations’ military industries. Furthermore, a working group was constituted to facilitate discussion in the sphere of counter-terrorism. During his visit, the Greek PM underscored Greece’s commitment to participate in the India-Middle East-Europe Economic Corridor (IMEC). 

This visit is the latest example of the various shifts in India’s foreign policy, especially towards the Mediterranean region. Given Greece and its neighboring regions’ strategic relevance in today’s geopolitics, this visit also forwards India’s strategic interests. Additionally, it comes amidst the conflicts in the Middle East and other regions, and the supply chain disruptions it affected. Experts claim that this rapid Indo-Greek friend-shoring is indicative of India’s larger strategy to respond to the rift in relations with Turkey. This is in the backdrop of a growing Turkey-Pakistan nexus, which has also fuelled Ankara’s belligerence against India’s aspirations and trade expansion plans in the Mediterranean.

A Few Good Reads

• Amidst the surge of quick commerce, Nesrine Malik unveils the hidden economics of exploitation in delivery apps.

• Harish Khare devises a political tool-kit on securing democracy and strengthening political opposition

• Afghanistan under the Taliban is the theatre of “new” geopolitical contestation amongst gulf powers, claims Kabir Taneja

• The article navigates the dilemma of curbing inflation while nurturing economic growth amid changing demand patterns.

• Amid the UNRWA controversy, Arkoprabho Hazra elucidates India's delicate balancing act between supporting humanitarian aid and opposing terrorism.