At the same time, machine learning poses unique dangers – with the power to turn governments and private industries into surveillance states that weaponize social media. artificial “intelligence” is only as smart as the quality of the underpinning data. There is enormous potential for AI learning itself to be skewed, and for “black boxes” to exist because of the absence of the collection of certain types of information needed to draw a complete and accurate picture.

At the digital frontier, a digital divide is already evident. Most AI research is being conducted in countries with certain shared characteristics: high incomes, high levels of education, large populations, good infrastructure and internet access, and conditions that foster entrepreneurship. Most data are also concentrated among a handful of countries and corporations.  These pioneers have “the first-mover advantage”, giving them an edge – when it comes to data, digital infrastructure, capital and their ability to set the terms for others to engage in related governance and ethical debates, and research agendas. Inequalities – evident in so many aspects of life – are already an issue in AI; the most-recent Government AI Readiness Index by Oxford Insights lays bare the divide: the average readiness score for North America and western Europe is more than twice that of sub-Saharan Africa and central and south Asia – the two lowest-ranked regions. Despite calls for inclusive AI practices,  power dynamics on new technologies sound like those of the old ones – raising concerns a lack of inclusion of Global South researchers into governance issues, and the potential for extractive and exploitative practices to arise.