Voices and Models: Can generative AI strengthen democracy and development?

We live in an era of fragmented geopolitics, polarized debates, and declining trust in democratic institutions. Against this backdrop, the rise of artificial intelligence (AI), especially generative AI, poses a crucial question: will it undermine democracy and deepen inequality, or can it be harnessed to build more inclusive, resilient, and responsive societies? Perhaps the answer is both. In a recent lecture at the University of Pisa, I explored this tension. AI is neither a silver bullet nor an inevitable threat. While it can be a powerful enabler, its impact ought to be shaped not by technology alone, but by governance, ethics, and the voices that define its trajectory.

AI and democracy

Democracy depends on the ability of citizens to deliberate, weigh evidence, and hold leaders to account. However, today’s high-choice information environment, amplified by algorithms, often undermines truth-seeking and deepens polarization. Generative AI adds new risks with the spread of deepfakes and synthetic disinformation. But AI can also strengthen democracy. Machine learning models detect corruption and track public budgets. Chatbots and local-language digital assistants make civic engagement more inclusive. And fact-checking algorithms provide real-time verification of claims. The challenge is achieving balance. Without safeguards, AI risks entrenching surveillance and exclusion. With the right design, it may empower citizens and restore trust in democratic institutions.

AI and global development

AI is already reshaping global development across multiple sectors:

• Education: From SolarSPELL’s offline digital libraries to Kwame, an AI teaching assistant in West Africa, tools are helping bridge teacher shortages and personalize learning. Apps like Voiceitt support students with disabilities by translating non-standard speech into text.
• Health: CAD4TB uses AI to detect tuberculosis from chest X-rays in seconds, bringing diagnostics to underserved communities. Other apps (e.g. Ruby Health) detect anemia from fingernail images or provide snakebite recognition in rural areas where treatment is scarce.
• Social Protection: Togo’s Novissi platform used machine learning and mobile data to deliver targeted cash transfers during the pandemic, reaching thousands who lacked formal bank accounts.
• Energy and Infrastructure: AI-driven predictive maintenance in India’s solar farms has boosted efficiency by 15%, while fraud detection systems in Bihar have curbed electricity theft. In Africa, AI helps identify off-grid communities best suited for renewable mini-grids.
• Agriculture: Ghana’s Farmerline created Darli AI, which provides farmers with regenerative farming advice and tracks yield improvements. In Malawi, the Ulangizi chatbot offers real-time guidance in local languages, supporting resilience after climate shocks.
• Climate Adaptation: The UN’s Early Warnings for All initiative uses AI and satellite imagery in Ethiopia to identify communities most vulnerable to climate disasters.

These examples show how AI can “leapfrog” traditional barriers, expanding access to education, healthcare, and energy. However, they also highlight risks: data gaps, bias, and the exclusion of those without digital access.

Governance and inequality

The governance of AI is deeply unequal. While ethical debates are dominated by Euro-American perspectives, the lived experiences of the Global South remain underrepresented. This risks imposing external standards that fail to reflect local realities. Moreover, digital divides, algorithmic bias, and linguistic marginalization threaten to widen inequalities. Most large AI models perform well in English but poorly in African and Indigenous languages, excluding millions from their benefits. Indeed, AI can empower but only when governance frameworks prioritize inclusion, transparency, and accountability. Partnerships across governments, academia, civil society, and business are essential. So too are investments in local innovation and research ecosystems.

Shaping plural futures

AI’s impact will not be decided by technology alone but by narratives and power. Who sets the agenda, whose voices are amplified, and which models prevail will determine whether AI strengthens democracy and development or undermines them. Going forward, we must:

• Change narratives to focus on agency, not inevitability.
• Amplify diverse voices, especially from the Global South, women, and marginalized communities.
• Adopt adaptive governance models that are inclusive, iterative, and responsive.

AI is a tool we can shape. The question is whether we choose futures that empower, include, and dignify or futures that entrench inequality and exclusion.

Based on a talk given at the Circle U. summer school on “Human-Centered, Inclusive, and Equitable Artificial Intelligence” at the University of Pisa, 17 September 2025.