Regulating artificial intelligence: Why one size won't fit all, say experts

Sectoral approach to AI rules allows greater flexibility to regulators, say experts

Alphabet CEO Sundar Pichai’s recent call for regulating artificial intelligence (AI) has reignited the debate on the matter. While Pichai is not the first corporate leader to do so — Tesla’s Elon Musk has been advocating this for the past several years — the timing is important.

AI-based applications, such as facial recognition, are increasingly finding their way into routine usage, and they are evolving. Given the ever-changing nature of AI technology, experts are of the view that rules to regulate it have to be application-based.

Sajai Singh, partner at law firm J Sagar Associates, states the dilemma that regulators across the globe face when it comes to regulating AI.

“What options on regulations do we have? Open, universal and transparent standardised set of regulation for the operation of machine learning (ML)/AI models? An agreement between all the concerned parties?” he asks. ML/AI regulation may be impossible to achieve without controlling ML/AI, which defeats its purpose, he says.

Salman Waris, managing partner, TechLegis, agrees and says horizontal regulations on AI tech will limit innovation, making it difficult for the law to adapt to the technology changes. “A sectoral approach to regulating AI allows greater flexibility, better implementation, and a targeted approach.”

Experts also point to the need for developing a regulatory framework around AI in India with the rise of the start-up ecosystem delivering AI products. But, most are against introducing any new standards in this regard to the proposed data protection Bill, asserting that privacy is only one aspect that affects AI. As Huzefa Tavawalla, head of disruptive technologies practice at Nishith Desai Associates, puts it: “The Data Protection Bill should be a broad, macro framework. Later on, we can enact industry-specific laws — from robotics to drones, to health care”.

He also stresses the need for deciding accountability parameters and AI rights. “Saudi Arabia has declared humanoid robot Sophia its citizen and granted her rights, but that is a one-off case. The question remains: Who owns AI, the creator of the algorithm or its operator?”

Experts note that corporate bodies have been granted legal rights and are accountable to their stakeholders, and the same principle can be applied to AI. But there is a catch: Corporate bodies aren’t truly independent as they are run by humans; AI can be truly independent.

So what are the dos and don’ts as governments across the world, including India, look to regulate AI. “Regulations should not be unnecessarily penal in nature, as that will have a chilling effect on innovation. Regulations need to allow companies to innovate, excusing innocent mistakes and only punishing reckless or willful violations,” says Rahul Matthan, partner, Trilegal.

The Indian government has been conscious of tapping the potential uses of AI. Piyush Goyal, while presenting the interim Budget last year, had announced the government was considering setting up a national centre for AI. The NITI Aayog, too, in its 2018 discussion paper stated: “In order for India to ride the AI innovation wave, a robust intellectual property framework is required.”
 

Some experts want global consensus on the AI framework and suggest a common point of interaction between regulatory authorities in the research and development area. Many believe that the laws should be based on the OECD Fair Information Practice Principles (which calls for AI systems being designed in a way that respects rule of law, human rights, democratic values, and diversity) and they should implement privacy by design.

There is another side to the conundrum: Governments across the globe have used/abused facial recognition technologies. Being a stakeholder, how impartial would they be in framing regulations?  “The only effective way to manage the use of technology by a government is for the government to proactively manage this use itself. What is needed today is a government initiative to regulate the proper use of facial recognition technology,” argues Waris.

Experts believe a broad regulatory framework will not only benefit end-consumers, but also facilitate large-scale adoption of AI tech by industry.