What does it look like to develop AI policies in Africa?
Tags
Policy-makers
AI Policy
Data Protection
Considerations for AI policy making in the Global South
When reviewing best practice for AI policy making in Africa, it’s critical to consider local, contextual issues likely to materialise and hinder equitable and responsible AI across the continent. This analysis looks at developing AI policies in Africa through four key lenses: data governance, access to infrastructure, data provision and strategic investments.
Data Governance
The first starting point for AI policy making in Africa should be through personal data protection. As African states roll out new services utilising AI-based systems, the protection of personal data in the development and deployment of these systems is vital.
Today, the majority of African countries have adopted a data protection law. However, these laws have developed different standards of protection, which makes ensuring interoperability - that despite having different provisions in place, a set of laws can function together - difficult. This is especially problematic considering these laws often restrict the transfer of personal data to other countries which do not have similar adequate data protection standards in place, thereby restricting the flow of data for AI development. The protection of personal data should fall under the broader strategic priority of African states to develop a data governance framework that addresses key issues on access to relevant and quality data for responsible AI development.
Data governance is not just about protection, but also access and interoperability, security and trust. Consequently, an important policy area for African governments to look at concerns the free flow of data. The cumulative effect of the adequacy requirement in African laws - that is, that personal data can only be shared across borders where an adequate standard of protection can be assured in the receiving country - is the restriction of free flow of data among countries. Additionally, other countries such as Nigeria have adopted various forms of light touch data localisation in sectors such as the telecommunication industry. Data localisation involves what are oftentimes protective measures to ensure data is protected from crossing country borders. This trend needs to be reversed and AI policies should explicitly promote cross border data flows in line with the African Union Data Policy Framework, which seeks to foster and facilitate cross border data flows as well as ensure that data are used in a sustainable manner that benefits society as a whole and does not harm people’s privacy, dignity and security.
Equitable access to infrastructure
Another key consideration in the development of African AI policies is establishing infrastructure that will enable the development of responsible AI systems. Two primary infrastructure needs are access to:
Reliable and affordable energy
Affordable internet
The twin problem of lack of reliable and affordable access to energy and internet creates a digital divide that policies should aim to bridge to drive equitable development and access to AI technologies and infrastructure across urban and rural areas. Further, data infrastructure such as cloud servers and warehouses may require specific software associated policies, which though technical, should be rooted in considerations of open software for facilitating public use of technology.
African Governments as a provider of data for AI
Against the backdrop of competition and intellectual property policy, a key factor to consider would be African governments as a provider of data for the development of AI.
African AI policies need to address the inequality between home grown start-ups and foreign based firms who have the capital to stifle domestic competition. An equitable policy approach would reimagine competition and intellectual property policy approaches that ensure infant tech industries enjoy a measure of protection and prevent abuse of dominant positions by foreign firms. Governments hold arguably the largest databases in the world and facilitating access to these datasets can open new frontiers for the development of AI-based systems. Governments have the potential and ability to drive AI development by using the data they hold to pursue the development of responsible AI.
Two primary ways through which government owned data could be provided to AI developers include:
Open data: the most well-known process for the government to share data
Data stewardship: an option for governments to facilitate direct access to data from data subjects to third party AI developers
An enabling policy landscape is required to achieve these. Under data stewardship, governments can serve as data institutions that steward data use towards public interest in various ways including protecting sensitive data and granting access under restricted conditions. Linked to these is a policy ecosystem that promotes interoperability and established data standards to facilitate access to data.
Afrocentric datasets and public-private partnerships
Strategic investments in Afrocentric datasets and specifically, low resource language corpora are also necessary for equitable and responsible AI in Africa. There are several use cases emerging in Africa of private sector investments in the development of databases in African languages to advance generative AI. For example, in Nigeria, the government has partnered with the private sector, Awarri, to launch Nigeria’s first multilingual large language model (LLM). The data captured will enable the LLM to be trained with five indigenous Nigerian languages. A policy landscape which encourages collaborations such as this can facilitate AI innovation that is relevant for local contexts.
Insights to impact
Consider protection of personal data in the development and deployment of AI-based systems and the free flow of data.
Establish policies that address the digital divide, such as reliable and affordable access to energy and the internet.
Facilitate access to data through open data or data stewardship.
Invest in Afrocentric datasets and look to partnerships in the private sector who can support this mission.