State of Internet Freedom in Africa 2025: Navigating the Implications of AI on Digital Democracy in Africa

State of Internet Freedom in Africa 2025: Navigating the Implications of AI on Digital Democracy in Africa is the 12th annual State of Internet Freedom in Africa (SIFA) report produced by CIPESA (Collaboration on International ICT Policy for East and Southern Africa), launched on 29 September 2025 at the Forum on Internet Freedom in Africa (FIFAfrica25) in Windhoek, Namibia. The 2025 edition is the first in the 12-year SIFA series to make artificial intelligence its central subject — a 14-country qualitative survey of AI's implications for digital democracy, civic space, and human rights across the African continent, authored by Juliet Nanfuka.

The SIFA report series: 2014–2025

CIPESA has published the SIFA report annually since 2014, with each edition examining a distinct dimension of digital rights across Africa and each edition launched at FIFAfrica — CIPESA's annual multi-stakeholder forum convening policymakers, journalists, platform operators, telecoms, civil society, and academics. The thematic arc of the twelve editions traces the shifting frontier of digital rights concerns on the continent: 2014 began with government censorship practices in East Africa; 2015 examined user experiences of privacy and security; 2016 mapped governmental strategies to restrict digital rights; 2017 focused on internet intermediaries; 2018 documented data protection gaps; 2019 reviewed two decades of internet control evolution; 2020 examined COVID-19's effects on digital rights and internet shutdowns; 2021 surveyed surveillance's chilling effects on democratic participation; 2022 documented the rise of biometric surveillance risks. The 2024 edition — Technology in African Elections — examined the interplay between technology and the 20+ African electoral contests that year, with AI-generated deepfakes documented in Rwanda and South Africa and a finding that few African countries had adopted AI in official election processes. The 2024 edition's AI-in-elections findings are the immediate precursor to the 2025 edition's expansion of that AI frame to the full democratic and civic-space picture.

Scope and methodology

The 2025 report is a qualitative study drawing on literature review and key informant interviews across 14 countries: Cameroon, Egypt, Ethiopia, Ghana, Kenya, Mozambique, Namibia, Nigeria, Rwanda, Senegal, South Africa, Tunisia, Uganda, and Zimbabwe. The country selection spans sub-Saharan, Francophone West African, North African, and Lusophone African contexts, covering a substantially broader cross-section than the 7-country East and Southern African focus of the Paradigm Initiative-led Legal Analysis of Laws, Policies and Government Strategies Relating to Artificial Intelligence published in November 2024.

Key findings

The report's organising analytical frame is a four-pillar model explaining AI's democratic trajectory in each country: a country's democratic credentials, economic gaps, legacy governance structures, and in-built AI design exclusions together determine whether AI serves or undermines civic participation. The correlation between democratic health and positive AI outcome is the report's central empirical claim.

Legislative gap across all 14 countries. None of the 14 study countries has AI-specific legislation, with data protection, cybercrime, and copyright instruments stretched to cover AI but found inadequate to govern its deployment. The report treats this as the primary structural risk to rights-respecting AI governance on the continent.

Democratic quality predicts AI deployment posture. The report finds that strong democracies — South Africa, Ghana, Namibia, and Senegal — are more likely to set positive standards for AI rollout, while countries with lower democratic credentials — Cameroon, Egypt, Ethiopia, and Rwanda — risk deploying AI to amplify digital authoritarianism and political repression. The political environment is the crucial determinant of AI's trajectory.

Surveillance and self-censorship. AI-powered surveillance systems are documented tracking, monitoring, and suppressing online voices across multiple study countries, leading to privacy violations and self-censorship — extending the pattern the 2021 SIFA edition first documented in the non-AI surveillance context.

State AI use cases. Positive government applications documented include South Africa's South African Revenue Service (SARS) AI assistant for tax assessments, Kenya's Sauti ya Bajeti platform for fiscal transparency, and Rwanda's deployment of AI for judicial system improvements. The report notes that high-democracy contexts are disproportionately represented in the positive use-case register.

Synthetic media and disinformation. Pro-government propagandists in Rwanda used large language models to mass-produce synthetic messages on social media — the first documented African case of LLM-enabled state-adjacent influence operation in the SIFA series.

African-language exclusion. The absence of large language models for indigenous African languages represents both a technical limitation and a symptom of broader technological exclusion that could deepen existing inequalities, positioning the dominant AI architectures as structurally excluding African language communities from participation in and benefit from AI systems.

Civil society response at FIFAfrica25

The report's launch at FIFAfrica25 prompted civil society participants to consolidate three priority recommendations: establishing human rights-centered AI governance frameworks across African jurisdictions; building local capacity for AI innovation and oversight; and developing AI systems in African languages as a structural correction to the exclusion the report documents. CIPESA, as a Global Network Initiative member, convened FIFAfrica25 in Windhoek — the 12th edition of an annual forum that has rotated across Kampala, Johannesburg, Accra, Addis Ababa, Lusaka, Dar es Salaam, Dakar, and Windhoek, drawing multi-stakeholder participation in rights-based responses to digital governance challenges. The forum convening is itself part of the civil society infrastructure through which the SIFA report's findings are disseminated and debated across the continent.

Posture within the corpus

This is the first CIPESA-authored publication entry in the corpus. CIPESA is already a recurring reference across the corpus: Daniel Mwesigwa of CIPESA Uganda presented on AI-enabled surveillance at the GISWatch 2019 IGF Berlin launch panel; CIPESA co-organised the February 2024 Dakar content moderation workshop with AfricTivistes during the Senegalese electoral crisis; and multiple funder entries name CIPESA as a grantee across the digital-rights funding ecology. The 2025 SIFA entry installs CIPESA's anchor publication in the graph, giving the recurring CIPESA references a publication to point toward.

Within the African civil society AI publications cluster, the 2025 SIFA sits alongside the Paradigm Initiative–led Legal Analysis of Laws, Policies and Government Strategies Relating to Artificial Intelligence in Kenya, Mauritius, Rwanda, South Sudan, Tanzania, Uganda and Zambia (7 countries, legal-analysis methodology, Anglophone East and Southern Africa) and the 2019 GISWatch AI edition (40 country reports, multi-author edited volume, APC and ARTICLE 19 co-publication). Where GISWatch 2019 is the continental multi-country survey of AI's human-rights implications and the Paradigm Initiative 2024 report is a focused legal analysis of East and Southern African AI governance frameworks, the CIPESA 2025 report occupies a third register: a single-authored qualitative political-economy survey of AI's democratic implications across a 14-country cross-section that spans all of Africa's major subregional and linguistic zones. The three publications together constitute the corpus's African civil society AI evidence base, each approaching the same continental challenge from a distinct methodological and thematic angle.