Regulating AI or "Regulating for AI"

The industry, Governments, and regulators these days are in a huddle to match the pace of AI developments and deployment. Understandably, there is a need too for that. It has taken less than a year for a focus shift from ‘Why’ AI is important to ‘How’ AI needs to be ‘THE’ strategy for businesses and nations alike. Summarizing the regulators’ dilemma leads to many questions including.

• At the first level, is there a need to regulate AI?
• Will AI regulation hinder innovation?
• What is required to be regulated in AI?
• What should be the approach- An AI law or guidelines (mandatory or voluntary)?
• Is an overarching AI law needed or different impact areas such as IP, privacy etc. can be managed through separate regulations?
• Is AI development alone that is required to be regulated or AI deployment too needs regulation?
• Given the data flows, how can an approach on internationally applicable regulation be worked out? How can the reciprocal behavior on AI regulations be handled?
• How to account for the cost of compliance of regulations on AI? Will it hinder FDI and the associated gains for countries?
• How are the existing digital regulations required to be changed to cater to AI requirements.
• How to ensure an equitable AI ecosystem bridging the digital divide?

Consequently, we have seen fragmented approaches to address the above. EU has taken the lead in coming out with the first AI Act almost repeating the GDPR story. In Asia, Japan took a step approach in coming out with guidelines on copyright, business AI etc. and have now announced an intent of a law. Singapore came out with first model governance framework covering generative AI and is the first nation to take AI as a ‘necessity’. India adopted the ‘AI for All’ approach making AI a mission-based program. South Korea is amending the existing digital regulations to make those AI ready. Australia took a consultative approach on white papers on ethical AI and responsible AI to start with.

Almost all the above are approaches to regulate AI. Given the rapid deployment of AI in all sectors, there is a need to move to ‘regulations for AI’ in sectors. The next wave of AI will not be a verticalized development centric but horizontal and deployment centric. In addition to the digital regulators preparing the groundwork for AI regulation, the sectoral experts and regulators need to work on the respective sectoral regulations and guidelines to make them AI ready. Some examples:

• The transport and motor vehicle laws and regulations to get ready for autonomous cars i.e. safety standards and for AI in traffic management.
• Civil aviation rules and guidelines to get ready for drones and also for using AI for aviation safety.
• Consumer laws to get ready for biases out of AI recommendations for products.
• Health systems related regulations to get ready on AI based diagnosis and treatment protocols e.g. regulating AI based medical devices and privacy etc.
• Agriculture related guidelines to get ready for precision AI based farming and privacy in smart farming.
• Financial regulations to cover AI in financial services and the algorithmic bias.

The above can only be addressed and governed through a joint approach between industry, sectoral and digital regulators. We have already seen some developments and references in this direction. The United Kingdom’s Financial Conduct Authority (FCA) has worked on the use of AI in financial services. The International Civil Aviation Organization (ICAO) has been holding discussions on developing standards and recommended practices (SARPs) for AI in aviation. The European Union’s Common Agricultural Policy (CAP) is incorporating AI-driven technologies to enhance precision agriculture, which includes the use of AI for crop monitoring, yield prediction, and resource management. In Australia, the government is working on a framework to address data privacy concerns in smart farming. The U.S. Food and Drug Administration (FDA) is developing a framework for AI-driven medical devices, emphasizing continuous monitoring, post-market surveillance, and real-world evidence collection. The EU’s Automated Driving System (ADS) Regulation sets stringent safety and testing requirements for self-driving cars, focusing on risk assessment, data logging, and cybersecurity. Singapore is updating its transport regulations to include standards for AI-based systems, ensuring they meet safety, efficiency, and transparency criteria.

Given the steep rise of AI on the development curve, this is the RIGHT time to start setting the house to avoid last minute scramble for creation of the right sectoral policy and regulatory ecosystems. There is a need to supplement the current approach of regulating AI with sectoral regulations getting ready for AI to address the readiness requirement due to shrinking time gap between AI development and deployment. In the end, it should not happen that technically the AI storm reaches home without sectoral readiness and preparedness. This will help all stakeholder groups get ready to reap benefits uniformly and holistically.

In summary, four lines generated by generative AI:

AI stands poised with promise, yet sectors seem to be running behind,

Technical readiness shines, but gaps in plans unwind,

Without the groundwork laid, how can progress find its way?

A canvas half-prepared may leave much hope astray.

The Views are personal