Work in Progress in Computer Vision and AI Surveillance in Africa

Computer vision is at first glance simply technology with general applications so that the political implications are not immediately apparent, unlike technologies such as bio-identity in which purpose is fused with the technology. The political valence of computer vision depends on the particular purposes for which computer vision is deployed. Although computer vision has in common with other AI technologies issues in the creation, representativity and access to datasets, not least the conditions of labour required inputs to AI systems, it has its own concerns. One area where computer vision has significant implications is for practises of surveillance. Enabling continuous visual surveillance at a scale hitherto unimaginable, computer vision (CV) is reshaping the surveillance landscape in Africa. Globally computer vision has been a topic of increasing interest and study due to its wide use in facial recognition systems, surveillance networks, virtual reality, and other similar fields. The full implications of widespread, interconnected, autonomous surveillance is not yet well understood, but they implicate human rights, particularly, the rights to privacy and autonomy. Artificial Intelligence technology has its origins in efforts to mimic human intelligence through the hardware and software of computer systems. While this goal remains elusive some AI technologies seek to emulate human abilities. Computer vision is one of several multi-purpose technologies developed under the broader category of AI. Similar multi-purpose technologies include natural language processing and decisions intelligence. As a multipurpose technology computer vision can be deployed for a number of different uses including analysing satellite or drone footage, facial recognition, gait recognition, number plate recognition and automated visual surveillance. Many African countries are increasingly introducing advanced AI surveillance tools and technologies to monitor, track and surveil their citizens, and cities are tapping into AI technology to monitor traffic and fight crime. Yet these technologies are deployed to accomplish a range of policy objectives— many are lawful, whereas some have an unclear mandate and hidden from public scrutiny. This research illustrates that analysing the implications of a particular technology for development is in many ways less revealing than to examining a technological phenomenon that has fused with its raison de'etre such as bio-id. However, a number of conclusions can be drawn from these two case studies: First of these is that the existing legal and governance frameworks for the use of computer vision for automated surveillance are inadequate. The use of facial recognition technologies can be expected to increase steadily, especially in the realm of public space surveillance. It is also possible that many countries in Africa with weak regulation can become testing grounds for emerging biometric technologies. Since governance is inadequate in constitutional democracies such as Botswana and South Africa it seems extremely unlikely that they are adequate elsewhere in Africa and our research does not suggest otherwise. Legal and governance frameworks are inadequate in at least three respects. They lack clarity as to how the right to privacy extends to being observed in a public space. They don't require transparency or reporting on the uses or possible abuses of automated surveillance. They also fail to impose clear duties on private profit driven actors using public spaces and providing public goods. The one exception to this is personal data protection regimes, although that in Botswana is not yet operational and in South Africa it is not yet fully operational.