ONE cannot deny the prominence artificial intelligence has gained with innovative uses in the political sphere — such as using AI to clone the voice of Imran Khan, who addressed a virtual PTI jalsa from Adiala jail. Activist Usama Khilji recognises the possible use of AI to circumvent political crackdowns but also warns of the “potential abuse” of AI technology, which could lead to misinformation and manipulation of elections. Foreign Secretary Syrus Qazi has previously stated that Pakistan recognises the “dual nature” of AI: while equitable and fair access to these new technologies is encouraged, there is concern regarding AI’s unregulated military uses.

Internationally, AI commands the attention of both the public and private sectors — with almost daily updates in legislative instruments, knowledge, and content. Last Friday, 27 countries of the European Union unanimously approved the Artificial Intelligence Act. The EU AI Act is widely described as the world’s first comprehensive law that regulates AI based on its capacity to cause harm. Closer to home, China is noted by some to have initiated the global movement to regulate AI when it introduced its Generative AI guidelines last year.

So where does Pakistan stand on the AI front? The Ministry of IT & Telecom introduced a draft national AI policy in May 2023. The proposed policy is part of the ministry’s Digital Pakistan vision that aims to produce an “overarching developmental agenda” to promote the adoption of AI efficiently and responsibly. The draft aligns with certain global standards for AI, such as its ethical and responsible use. For instance, the ‘triple-helix’ partnership between the research community, industry and government is similar to Singapore’s national AI policy. Singapore was one of the first countries to introduce a national AI strategy.

While the draft AI policy sets the foundation for AI transformation in Pakistan, there is room for improvement. Take, for example, the lack of emphasis on ethical and transparent AI systems. In this regard, one can turn to the ‘Artificial Intelligence Risk Management Frame­work’ introduced by the US National Institute of Standards and Technology. Aiming to govern AI systems and mitigate risks, the framework promotes “trustworthy” AI systems, comprised of characteristics such as explainability, resilience, and data privacy.

Where does Pakistan stand on regulating AI?

An effective AI policy should ideally emphasise concepts of data quality and data privacy. Unfortunately, the cart seems to be placed before the horse. The AI policy proposes an AI directorate to be constituted under a ‘National Commission for Personal Data Protection’. However, the NCPDP is to be established under the Personal Data Protection Bill which is still pending approval by the National Assembly. While the draft AI policy encourages international and bilateral cooperation for AI development, the emphasis on ‘data localisation’ by the PDPB could hamper cross-border data flows.

Other proposals include the establishment of a robust regulatory framework, a ‘National AI Fund’, and ‘centres of excellence’ in AI that will serve as AI innovation hubs. Naturally, these endeavours cannot be viable unless sufficient infrastructure, resources and funds are available. Rather than being a passive regulator, the government of Pakistan will have to actively enable the effective implementation of the AI policy. For example, Singapore’s AI policy states that it had allocated 500 million Singaporean dollars through AI Singapore under the Research, Innovation and Enterprise 2020 and 2025 plans.

Any attempt at AI regulation needs to be more than a hurried response to a global trend and should consider the situation on the ground. The Digital Rights Foundation has stated that the draft AI policy should be human-rights centric. Additionally, the Islamabad Policy Research Institute found that there is “insufficient and controlled” availability of data, which can hamper implementation in different sectors (such as education, climate, and agriculture). The most recurring critique is the absence of any mention of diversity and inclusion.

The draft AI policy was to be finalised by December last year. However, like the PDPB, the proposed AI policy seems to be collecting dust. Contrary to international best practices, the draft AI policy was released without a period of consultation. Since then, critique from those at home and inspiration from those abroad can be used by the incoming government to regulate AI effectively. Hopefully, in its final version, the AI policy won’t just be — as journalist Mutaher Khan puts it — a piece of paper to “serve samosas”.

The writer is a lawyer working in data privacy and technology.

Published in Dawn, February 9th, 2024