AI and data governance are inseparable. Without data there’s no AI, and without high-quality, representative data, there’s no high-quality or representative AI.

AI can improve the management of public data. It allows large databases to be standardised and analysed at a speed previously impossible, which can detect irregularities in public procurement, improve government services and strengthen public policies.

But it also amplifies the risks. Given the enormous data-processing capacity of some private companies, the challenge lies in governance. The question is what can be done with the data and what cannot, and to what extent citizens have the power to decide on the use of their own information.

The Global Digital Compact, adopted by the United Nations (UN) General Assembly, is one of the most far-reaching instruments on digital policy that currently exists. It covers everything from connectivity and infrastructure to data governance and AI. Within this framework, two bodies were established: a working group on data governance, which I take part in as a civil society representative, and a panel of AI experts mandated to make recommendations to the UN. Both include academia, civil society, governments, international organisations and the private sector.

One of the central debates we encounter in these forums and in our research on regulatory frameworks revolves around how existing regulations on access to information, copyright and personal data privacy provide a basis for understanding what can and cannot be done in the development of AI. Building on this, the question is what specific AI regulations need to be developed, or what updates need to be made to existing ones.

The answer varies by country. In a recent report, we analysed how six countries from different regions — Chile, Kenya, Mauritius, Mexico, the Philippines and Taiwan — are addressing this question. We found that many already have frameworks that, if properly applied, can govern AI development. In many cases, adapting existing regulations may be more effective than building new ones from scratch.

At Open Data Charter, we advocate that, whatever path each country chooses, human rights must be at the centre. That’s the principle we take to international forums, where civil society organisations, governments and tech companies don’t always agree on what should be prioritised but sit down to debate collaboratively.

One of the main challenges is funding. It’s not enough to pay for a ticket to a summit. Meaningful participation requires preparatory work, such as conducting research, consulting with other organisations, reviewing documents and developing robust positions. That takes time and resources, and the funds available to civil society are dwindling, which directly impacts on the quality of participation.

A second challenge is the fragmentation of spaces. Today, there are multiple forums making decisions on data governance and AI at the same time. For small organisations, even those with considerable technical expertise, it is very difficult to keep meaningful track of them all.

As resources dwindle and forums multiply, the risk is that the voices of organisations that have been working for years on digital rights and human rights in the digital sphere will fail to be heard precisely when they are most needed.

We know we can’t influence international policy on our own. That’s why we work in partnership with other digital rights organisations and align positions in international forums, and work directly with governments on policy implementation. We also understand there’s no one-size-fits-all solution. The cultural, institutional and legal context of each country matters, and any process must begin with active listening.

Results sometimes come about in unexpected ways. In Chile, we worked with the Ministry of the Environment on opening environmental data. Some time later, we learnt that the new environmental transparency law had incorporated open formats as a legal requirement, something that those involved in drafting it attributed to that joint effort. It was the result not of a deliberate campaign, but of day-to-day work with officials who subsequently brought those ideas into the law.

Something similar happened with the Open Data Guide to Combat Corruption, which we developed together with the Mexican chapter of Transparency International. The guide identifies 30 high-value datasets for anti-corruption policy. The Organization of American States (OAS) used it as the basis for its Inter-American Open Data Program to Combat Corruption, approved in 2018, with over 10 implementations across the region. It was not us who took it to the OAS. It was an institutional proposal that the organisation simply adopted.