Submit your Article

Have you ever imagined a scenario where the National Police issue search warrants or execute arrests with the assistance of AI tools?

Would any type of ethical safeguards be respected in such a process? 

Would human control over the final decision of taking action, in the pursuit of justice, against an alleged perpetrator of a crime, be assured? 

In the EU AI Act, the first ever legislation governing the use of artificial intelligence entered into force in August 2024 and in the process of phase-wise implementation, there are clear provisions on when and in which circumstances, and under which type of oversight, the law enforcement agencies can make use of AI-enhanced tools. For example, AI-powered real-time facial recognition in public places is prohibited but there are exceptions for law enforcement agencies of the member-states. 

Indeed, the Act has come under fire for allowing too many loopholes for police use. But, in a trailblazing move, it has also put in place a risk-based approach with four levels of risks, including AI deployments with unacceptable risks that are forbidden. It is to be updated and revisited to ensure it will remain a “fit for purpose” instrument for a technology with apparently unlimited potential. 

This example of emerging AI regulations in the EU proves the complexities in finding a right balance not only on how to make the most effective use of the AI. It also shows the challenges of ensuring that ethical guardrails are in place before the deployment of new AI instruments whose full potential and capacities are still not fully comprehended. That’s why the recent news that the Nepal Police has not only acquired some AI software from India but also trained some of its staff on its use should be taken as very alarming. 

In a fast-evolving scenario of AI development where there have been a lot of talks about regulations but much less action, the recent government act of issuing a draft AI Policy is a positive step. The draft lists out several worthwhile policy objectives and it correctly highlights how any development and use of artificial intelligence in the country should always be focused on the creation of positive impacts for the wider society. Yet there are concerns about the lack of specificities and technicalities in relation to the implementation of any future AI regulations. At the same time, it is vital to contextualize any AI Policy with the urgent need for the country to have in place strong data protection and privacy laws, cybersecurity and digital infrastructure that Nepal still lacks.

Finalizing the AI Policy, therefore, would require, as suggested by the Digital Rights Nepal and AI Association Nepal, establishing a robust framework to ensure proper conduct in the way data protection and privacy rights are ensured by AI developers and their users. These represent major conundrums not only for Nepal but for the international community as a whole and are crucial themes widely underlined by the UN High Level Advisory Body on AI through its Governing AI for Humanity report.

One of the major deficiencies of the draft is its “generalized and listing all” approach without any timeline. For example, legislating a data protection framework would be itself a gigantic effort with a high degree of complexity. Addressing it head-on alone would, as a consequence, demand a strong sense of urgency on the part of the legislator and executive powers. But lack of specificities is not the only problem of the draft. 

A concerning aspect is also the system of governance that is envisioned to govern AI in the country. First of all, the document foresees the creation of an AI Regulatory Council whose chairperson would be the Minister for Communications and Information Technology. Together with other very high-ranking officials, including the Governor, this mechanism would, among others, issue ethical AI guidelines and standards.

The draft also envisions a National AI Center that, among other things, will be in charge of regulating the development and use of AI at the national level and overall coordination and evaluation of AI development.

It is crucial to reassess the purpose of establishing a high-level body like the AI Regulatory Council, as their effectiveness is often questionable, with many of such high-level bodies struggling to even meet the quorum for meetings. Instead, the focus should shift toward empowering an institution like the National AI Center, which could serve as the "guardian" of future AI legislation and ensure its proper implementation.

For example, the EU AI Act established a powerful European AI Office within the European Commission, giving it wide-ranging powers, including enforcement and implementation. But this is not the only mechanism created through European legislation. There are also a Scientific Panel, composed of independent experts in the field of AI and the Advisory Forum, representing a diverse selection of stakeholders. There is also a European Artificial Intelligence Board comprising representatives of member-states. This open, inclusive approach to governing AI development where together with policymakers, experts and members of the civil society have a seat on the table, is innovative.

At RightsCon 2025, held in Taiwan on 24-27 Feb 2025, AI discussions emphasized inclusivity, ethics and accountability in AI governance and development. Key sessions explored civil society’s role in AI policymaking, particularly in Asia and Latin America, and the integration of global perspectives for responsible generative AI. Topics like multilingual AI, neurotechnology governance and feminist AI highlighted the need for diverse voices in shaping equitable AI systems. Further, sessions on AI for climate action, healthcare and natural resource governance underscored its potential to address global challenges while ensuring rights-respecting approaches. The conference also discussed AI bias, fairness and the democratization of AI infrastructure, advocating for transparency and public participation. It is important for Nepal to consider these issues while finalizing the AI draft policy and legislating AI regulation. 

Moreover, recently, the Tony Blair Institute for Global Change has developed a new report, “How Leaders in the Global South Can Devise AI Regulations that Devise Innovation”, that should be thoroughly taken into consideration. 

If Nepal wants to really become an emerging IT hub as envisioned by the government, then it needs to step up its tempo on AI regulation. The country should have a holistic AI framework in place founded on strong data and privacy rights pillars. Together with strong ethical guardrails, Nepal can design an agile system that, while not over cumbersome, can ensure safety and AI alignment with human rights. 

By studying global practices and adapting them to local context, Nepal can have an ethically solid and innovation-promoting AI framework in place.

Any use of AI tools by state agencies, including law enforcement, should be stopped before any such a framework is established.