ICTworks

Artificial intelligence is often posited as a solution to the myriad difficulties faced on the African continent. AI discourse is littered with articles that proclaim an optimistic vision for AI-driven social change, with titles that suggest “the future is intelligent”, and that we must “harness AI in Africa.”

African Governments Using AI

We have harnessed it. In Togo, AI has been used to target the poorest cantons for social funds and in Zambia it is being used to identify and mitigate mis-and disinformation during elections. However, like all ‘silver bullet’ technologies, even a successful solution to one problem can create additional, different ones.

AI has reportedly been used in lethal autonomous weapons systems in Libya, and in an alleged rollout of national facial recognition systems in Zimbabwe, and a significant amount of research has documented how AI can entrench biases, deepen injustices, and infringe on a range of human rights.

This is the dichotomous nature of AI technologies. They have enormous potential for good, but also pose significant risks. In addition, the use of AI technologies is by no means confined to democratically elected states, where the citizenry might exercise control over harmful policies and decisions: the vast possible uses of AI technologies by businesses and private actors, often based in the global north, must also be accounted for.

Artificial Intelligence Governance

In response, states have grown increasingly interested in the governance of AI. According to the OECD there are over 700 AI policy initiatives that have been implemented by 60 countries, since 2017. Further, forty- two countries adopted the first set of intergovernmental policy guidelines on AI developed by the OECD in 2019. However, only 5 African countries contribute to the OECD’s 60-country membership, and none of the signatory countries were African.

Africa also performed poorly according to the Oxford Insights 2021 AI Readiness Index, which ranks 160 countries by how prepared their governments are to use AI in public services. The report reviewed 41 countries from the Sub-Saharan Africa region and found that the region’s average score is 31.61 out of 100, far below the global average score of 47 out 100.

So what measures are African countries taking in response to AI?

How African Countries Govern AI

AI Governance in Africa answers that question by mapping the current state of AI governance in Africa. We look at what instruments exist at the continental level and domestically. An analysis of the substance of the instruments is beyond the scope of this report but will be included in future reports.

• No country has dedicated AI legislation – though Mauritius has partial legislation on AI.
• 30 countries have data protection legislation that addresses automated decision making.
• 4 countries have a national AI strategy.
• 1 country has a draft policy or a white/green paper on AI.
• 13 countries have established an expert commission or taskforce on AI.
• 6 countries include AI as a priority in their National Development Plan, while 4 other countries’ plans make partial mention of AI.

These findings complement recent survey research conducted by UNESCO which finds that there is a “wide variation in the nature and scope of policy instruments used by countries in Africa for the governance of AI.” UNESCO found significant ongoing initiatives to govern the use of AI:

• Eighteen out of 32 countries have ongoing initiatives to guide the development of AI at the national level.
• The development and use of AI is a priority as per the national development plans in 21 out of 32 countries.
• Out of the 32 respondents, thirteen countries have launched AI strategies, thirteen have developed AI policies, six have reported enacting legislation to address some of the challenges of AI, twelve have established Centres of Excellence on AI, and three have reported issuing ethical guidelines for AI.
• According to research by UNESCO, the development and use of AI is listed as a priority in the national development plans in 21 out of the 32 countries in Africa surveyed.

Our research finds that many of AI policy initiatives in Africa are not yet in the public domain. This indicates either that the development of AI frameworks remains opaque or that progress has been slow in reaching final outputs, or a combination of both. This is deeply concerning and implies a lack of public participation in and oversight over the development of AI regulatory frameworks in countries in Africa.

A light synopsis of AI Governance in Africa by the Collaboration on International ICT Policy for East and Southern Africa