Speaking during the online Participatory AI Research & Practice Symposium Panel, Abdullahi warns that many AI governance systems across Africa are being built on digital participation models that assume widespread connectivity, despite persistent and significant digital access gaps across the continent.

“Participatory AI governance is often framed as a democratic process, but participation requires access and in context of digital inequity that access collapses and that requires different mechanisms,” she says in her presentation titled Rethinking Participatory AI Governance Under Digital Inequity.

 

“The problem here is that many AI governance processes rely on digital mechanisms,” she says. “There is an assumption that citizens can participate digitally through online portals, virtual consultations and web-based feedback platforms. But what happens when millions of people cannot connect?”

 

Her intervention comes as African governments intensify efforts to position themselves within the global artificial intelligence economy. Abdullahi cites McKinsey projections suggesting AI could contribute billions of dollars to Africa’s economy by 2030, with more than 15 African countries already having developed national AI strategies as of 2025.

However, she argues that these ambitions are unfolding against a structural constraint: widespread digital exclusion.

According to data presented at the symposium, 64% of Africans remain offline, while high data costs continue to deepen inequality, particularly in rural and underserved communities.

Digital exclusion threatens legitimacy of AI governance

Abdullahi says many AI governance frameworks rely heavily on online consultation mechanisms that automatically exclude large segments of the population.

“The problem here is that many AI governance processes rely on digital mechanisms,” she says. “There is an assumption that citizens can participate digitally through online portals, virtual consultations and web-based feedback platforms. But what happens when millions of people cannot connect?”

She argues that this structural disconnect raises fundamental questions about the legitimacy and inclusiveness of emerging AI governance systems across Africa.

“When baseline digital access is uneven, participatory legitimacy cannot be assumed,” she says.

Governance framework analysis reveals inclusion gaps

To assess the issue, Abdullahi adapts Archon Fung’s Democracy Cube framework to evaluate AI governance models through the lens of digital inclusion. Her adapted model examines who participates, how participation occurs, and what level of influence participants have on policy outcomes, while also accounting for infrastructure access, affordability, language barriers, and digital literacy.

She applies the framework to three major policy initiatives: Nigeria’s National Artificial Intelligence Strategy, Kenya’s Artificial Intelligence Strategy 2025–2030, and the African Union Continental Artificial Intelligence Strategy.

The findings highlight varying levels of inclusivity across the three governance models.

For Nigeria, Abdullahi notes that while the strategy acknowledges digital inequality and infrastructure gaps, the consultation process remains heavily dependent on digital participation channels.

She says Nigeria’s AI strategy development engaged “over 120 internal and external experts,” but argues that this approach risks excluding a significant portion of the population, including the estimated 55% of Nigerians who remain offline.

“Nigeria utilised an in-person workshop and then followed with an online portal for public review,” she says. “There were no primary offline mechanisms for the public to participate.”

She also highlights linguistic exclusion challenges in Nigeria’s consultation process.

“For a country like Nigeria, where I’m from actually, that has over 500 languages, that is missing a key multilingual approach,” she says, noting that engagement was conducted primarily in English.

African Union, Kenya show contrasting approaches

The African Union Continental AI Strategy, she notes, follows a largely expert-driven model anchored in institutional and technical working groups.

“The AU takes a more expert-only approach, relying heavily on specialized task forces and institutional experts,” she says.

While the AU framework references community-oriented principles, Abdullahi argues that it lacks clear mechanisms to track or integrate input from digitally marginalised populations.

By contrast, Kenya emerges as the most inclusive of the three case studies.

According to her analysis, Kenya conducted offline town hall meetings across 17 counties and incorporated Swahili-first AI considerations within its policy framework.

“Kenya demonstrated a stronger commitment to linguistic and physical accessibility,” she says.

However, she notes that limitations persist, as many consultations were still concentrated in urban innovation hubs and conducted predominantly in English.

Abdullahi argues that a broader structural issue runs through all three policy frameworks: digital infrastructure is primarily treated as an economic development enabler rather than a democratic governance requirement.

“Across all three of them, digital infrastructure is identified and framed in the strategies as an AI development prerequisite, but not as an AI governance prerequisite,” she says.

She warns that this framing risks widening existing inequalities as governments expand AI deployment across critical sectors including public services, healthcare, education, finance, and security.

“When we do not have the full consideration of digitally excluded individuals, the risk here is that as we’re advancing AI development and other advanced technologies, we risk widening the digital divide,” she says.

Call for offline-first AI governance models

To address these challenges, Abdullahi calls for the deliberate integration of offline and intermediary participation mechanisms into AI governance systems, rather than treating them as supplementary measures.

“It’s a necessity to embed offline and intermediary mechanisms alongside digital platforms,” she says. “But it should not be an afterthought, but a part of the actual core design.”

She also urges policymakers to clearly demonstrate how citizen input, particularly from marginalised groups, directly influences final policy outcomes.

“So showing that they actually had influence, not just that there was input and consultation from them, but reflecting clearly how that impacted the outcome,” she says.

No one-size-fits-all approach for Africa’s AI governance

The presentation further cautions against uniform AI governance models across Africa, citing the continent’s deep linguistic, cultural, and socioeconomic diversity.

“We can’t have a one-size-fits-all approach across all countries,” she says. “Solutions cannot be identical everywhere.”

As African nations accelerate AI strategy development and compete for investment in emerging technologies, the research underscores a critical governance question: whether the citizens most affected by AI systems are meaningfully included in shaping the rules that govern them.