TLDR: A research paper by African experts outlines the continent’s unique AI safety risks, from deepfake-fueled election interference and digital colonialism to job displacement and environmental impact. It highlights the current gaps in African AI governance and proposes a five-point action plan: a human rights-based policy, establishing an African AI Safety Institute, boosting public AI literacy, developing multilingual early warning systems, and an annual AU-level AI Safety & Security Forum to ensure safe and equitable AI development in Africa and integrate African voices into global AI governance.
Artificial intelligence (AI) is rapidly transforming economies, societies, and environments worldwide, and Africa is no exception. However, a recent research paper, authored by a diverse group of researchers and practitioners from across the continent, highlights that African nations and their people face distinct AI safety risks that are often overlooked in global discussions. The paper, titled Toward an African Agenda for AI Safety, emphasizes the urgent need for African perspectives to be integrated into the emerging global AI safety governance agenda.
Understanding Africa’s Unique AI Risk Profile
The paper categorizes AI risks into three broad areas: malicious use, malfunction, and systemic risks, detailing how each uniquely impacts the African context.
Risks from Malicious Use: AI technologies can be misused to undermine democracy and human rights. Examples include the potential for facial recognition to be used for surveillance by authoritarian regimes, as feared by activists in Zimbabwe and Uganda. AI-generated content, such as deepfakes and disinformation, poses a significant threat to democratic processes, especially in countries with fragile institutions and high social media penetration. During Kenya’s 2022 elections, AI-altered audio clips and social media bots spread false narratives, exacerbating ethnic tensions. The manipulation of public opinion through AI-generated deepfakes can also have disastrous health and safety impacts, as seen during the COVID-19 pandemic with disinformation about treatments. Furthermore, the paper warns of external influence, where foreign tech giants, particularly from China and the US, expand their digital infrastructure in Africa, raising concerns about technological dependency and digital colonialism. The militarization of AI, including lethal autonomous weapons and the use of AI by private military contractors or terrorist groups, also presents a grave concern, especially given existing conflicts and arms trafficking in regions like the Sahel. Finally, AI contributes to cybercrime and financial malpractice, with AI-powered phishing and voice cloning targeting Africa’s booming fintech sector, and AI-assisted programming tools potentially introducing vulnerabilities into digital systems. The rise of online gender-based violence, including deepfakes and algorithmic amplification of misogyny, is another critical malicious use risk, disproportionately affecting women and girls.
Risks from Malfunction: The reliability and safety of AI systems in Africa are challenged by several factors. Many AI models are trained on Euro-American datasets, leading to inaccuracies and biases when applied to diverse African contexts. Infrastructure limitations, such as unreliable electricity and a lack of quality, structured data, further compromise AI reliability, especially in critical applications like medical diagnostics. AI systems are also prone to ‘hallucinations,’ misrepresenting facts, which is particularly concerning for AI use in education. Algorithmic bias is a significant issue, with facial recognition software showing higher error rates for dark-skinned women and large language models often reflecting Western values, exacerbating historical inequalities in areas like maternal healthcare and loan financing.
Systemic Risks: AI poses unique economic and environmental risks. AI-guided automation threatens to disrupt traditional developmental pathways, with entry-level jobs being particularly susceptible. This is a major concern for African countries that have positioned themselves as business process outsourcing (BPO) destinations, as AI could make it more cost-effective for foreign firms to ‘de-couple’ operations. Women and gig workers are especially vulnerable to job displacement. Environmentally, the development and use of frontier AI models contribute to increased carbon dioxide emissions, energy consumption, and water usage for cooling data centers, which is particularly problematic for a continent already disproportionately burdened by climate change and water scarcity. The rapid turnover of specialized AI hardware also contributes to electronic waste (e-waste), posing significant health risks in informal recycling sectors, with only a few African countries having dedicated e-waste legislation.
The State of AI Governance in Africa
Despite these pressing risks, AI safety remains an emerging area of governance in Africa. The Global Index on Responsible AI indicates that Africa has the lowest average score in ‘Safety, Accuracy and Reliability,’ with only 11 out of 41 assessed countries showing concrete activity on safety. Kenya stands out as the only African country demonstrating efforts across government frameworks, actions, and non-state actors in AI safety. Other countries like Ghana, Morocco, and Rwanda are also making strides in developing policy frameworks that emphasize ethical and safe AI. The African Union (AU) has shown commitment through its Continental AI Strategy, which calls for regulatory instruments, an annual conference on AI Safety and Security, and an expert group to assess AI’s impact on peace and security. While many African countries have foundational laws like data protection and cybersecurity, there’s a considerable lag in developing specific AI regulatory legislation.
A Five-Point Action Plan for AI Safety in Africa
To address these challenges and ensure Africa’s agency in shaping the future of AI, the paper proposes a comprehensive five-pillar action plan:
1. A Human Rights-Based Approach to AI Safety:
This approach would embed human rights principles like non-discrimination, privacy, and freedom of expression into the design, development, and deployment of AI systems. It emphasizes meaningful community participation and tasks governments with developing strong legal frameworks to protect vulnerable populations from AI’s harmful socio-economic effects.
2. Establishment of a Dedicated African AI Safety Institute:
Leveraging Kenya’s involvement in the international AI Safety Institute Network, the paper proposes creating a continental institute. This institute would test locally developed AI models, conduct research on regional AI safety risks, develop policy innovations, and secure access to advanced computing resources for evaluating frontier AI models. It would serve as a hub for research, policy development, and capacity-building, ensuring African contexts are integrated into global AI safety conversations.
3. Increase Public Literacy on AI:
Addressing the existing skill and knowledge gap, this pillar advocates for capacity building among students, practitioners, and policymakers. Strategies include creating accessible educational resources (like MOOCs), integrating AI safety into curricula from K-12 to vocational training, and launching public awareness campaigns to foster critical thinking and media literacy, especially against AI-generated misinformation.
4. Develop Early Warning Systems for Africa with Inclusive Benchmarks for 25+ African Languages:
This crucial step involves building continent-wide early warning and monitoring systems to detect AI-driven threats like deepfakes and coordinated disinformation campaigns in real-time. The system would prioritize linguistic diversity, incorporating local languages and dialects into data collection and risk assessment, and producing open datasets and AI models for African languages. This would strengthen information integrity and resilience, empowering communities, journalists, and authorities to respond swiftly to harmful content.
Also Read:
- Understanding the Full Spectrum of AI Risks: A New Research Perspective
- The Imperative for a Unified Global AI Governance Framework
5. Annual AU-Level Forum on Safety and Security of AI in Africa:
Aligning with the AU’s Continental AI Strategy, this recommendation calls for a permanent annual forum engaging all 55 member states. The forum would serve as a high-level platform for knowledge exchange, harmonizing standards, producing an annual ‘State of AI Safety and Security in Africa Report,’ developing shared policy frameworks, negotiating with global technology providers, and liaising with international partners to manage transboundary AI threats.
By implementing this collective agenda, Africa can not only safeguard its communities and ecosystems from AI’s potential harms but also assert its agency in steering the future of AI toward equity, resilience, and shared prosperity, ensuring that AI development serves the continent’s unique needs and values.
