One example is the Observatory of Public Policies, a Tunisian civil society-led initiative to strengthen data-informed governance. It focuses on positive accountability, so it doesn’t seek to sanction institutions but to create structured transparency mechanisms that build trust between people and authorities. Its objective is to reframe public debate around verifiable evidence rather than ideological polarisation and equip civil society, media and researchers with tools to hold public institutions to account in a structured and credible way.

The initiative has shown that stronger public data ecosystems can shift governance cultures from political confrontation toward evidence-based policy dialogue. But it has also faced a major structural challenge: access to disaggregated, interoperable and reliable public data remains limited. Data is fragmented across ministries, inconsistently formatted or inaccessible. Essential datasets often lack gender and territorial breakdowns, making it hard to assess equity and inclusion.

These constraints reveal that the debate is not only about transparency, but about institutional capacity and data sovereignty. When public data infrastructures are weak, countries become dependent on external providers and proprietary platforms, limiting their ability to set standards, safeguard public-interest use of data and retain strategic control over AI-enabled decision-making. Data becomes a strategic asset extracted into global AI value chains, without corresponding local capacity to capture value, ensure accountability or shape norms.

Strengthening domestic data institutions is therefore a prerequisite for both democratic accountability and sovereign policymaking in the age of AI.

Collecting data isn’t just a technical exercise, it’s also a political and social process shaped by capacity, consent, power and trust. It can be sensitive, particularly when linked to security, surveillance or taxation.

In many global south countries, data is collected through AI-enabled tools, digital platforms or mobile services operated by companies based abroad. While states remain responsible for protecting citizens’ rights, they often have limited oversight over how this data is processed, stored and monetised.

This creates a pattern similar to extractive economic models: countries with limited public data capacity become major data providers in global AI value chains but have little influence over governance, standards or value creation.

AI governance therefore intersects with geopolitical positioning: it concerns who defines interoperability norms, who governs cross-border data flows and who ultimately benefits from the value generated by AI systems.

When algorithmic decisions are based on poor data, they become difficult to contest, particularly when citizens and governments do not fully understand how they are produced. When data systems are externally controlled, fragmented or opaque, civil society struggles to access information, scrutinise decision-making and challenge harmful outcomes.

AI outputs are often treated as authoritative or neutral, even when based on flawed or incomplete data. This marginalises alternative forms of knowledge produced by civil society, communities and local researchers.

For women’s rights organisations, the absence of gender-disaggregated data makes it harder to document discrimination and advocate for change. When AI systems are trained on such gaps, they reproduce gender bias at scale.

Regulatory models and technical standards developed in data-rich contexts are often promoted as global solutions. But if they don’t address who collects data, how it is governed and whose interests it serves, these governance frameworks risk overlooking local realities.

If AI is to support inclusive development and protect civic space, governance efforts must start with data justice. This means investing in public data systems, transparency around data sources, clear rules on data sharing, safeguards against exploitation and meaningful civil society participation.

Beyond regulatory alignment, this means consolidating AI initiatives from the global south, strengthening south-south cooperation and embedding AI-enabled public policies in locally defined priorities.

AI should not be an imported governance model. It should be a human-centred public policy tool grounded in evidence, guided by human rights principles and designed to enhance accountability, inclusion and institutional trust.

Without accountable, representative and trustworthy data systems, AI risks amplifying exclusion, inequality and invisibility, particularly where civic space is already under pressure.