How AI is upgrading African dictatorship

The IDS mapping shows cameras clustered where opposition parties organize, not where ordinary crime is highest Originally published on Global Voices Advox Illustration showing a dictator using surveillance and automated control to watch, predict, and suppress citizens. Illustration by Khalid Bencherif. Used with permission. By Khalid Bencherif “On coins, on stamps, on the covers of books, on banners, on posters and on the wrapping of a cigarette packet — everywhere. Always the eyes watching you and the voice enveloping you. Asleep or awake, working or eating, indoors or out of doors, in the bath or in bed — no escape. Nothing was your own except the few cubic centimetres inside yo ur skull.” – George Orwell, “Nineteen Eighty-Four.” Orwell could not have imagined that eventually, even those few cubic centimetres inside the skull would be contested on a continent already plagued by corruption and tyranny. This matters especially on a continent where corruption, impunity, and executive appetite long predate artificial intelligence. And where AI adoption is outpacing the development of rights-respecting legal frameworks Africa rushes for smart tools of tyranny In the past, tyranny in Africa required prisons, informants, secret police, and visible repression. Today, more and more of it arrives as software, financed by credit, wrapped in the language of modernisation, and sold as an upgrade to public safety. A March 2026 study by the Institute of Development Studies (IDS) and the African Digital Rights Network found that 11 African governments had collectively spent more than USD 2 billion on AI-powered surveillance systems. Nigeria alone accounted for more than USD 470 million. The kit includes high-definition CCTV, plate readers, facial recognition, biometric identity layers, and a control room where the feeds converge. Much of it is supplied or financed by Chinese firms and banks , while other layers come mainly from Israeli firms . Statistics about AI-powered surveillance. By the Institute of Development Studies. Used with permission. The common justification usually given by African governments is that acquiring this equipment is necessary to fight crime, but the crime numbers are not decreasing. IDS researchers, having examined the installations across several cities, found little evidence that the cameras reduce offending , and the evidence actually shows that cameras cluster in neighbourhoods where opposition parties organise, where protests have happened and where the press has “made trouble” for ruling regimes. The info-infrastructure of dictatorships in Africa is not new, and it won’t be starting from zero. In most African countries, the state has long had an appetite for data. Records from communications, tax rolls, bank accounts, and administrative registries have been collected for decades and used with little legal restraint, even where protections exist on paper. What is new is the engine. AI turns the dormant files into a queryable archive. A name, a network, a pattern of travel, a history of transactions, now surface in seconds. Even the KGB, the Soviet Union ’s secret police, as Yuval Noah Harari has pointed out , did not have enough people to read millions of reports on millions of citizens every day. The dust in the archive was itself a form of freedom. AI dissolves that dust. The AI-enabled surveillance now coming online in African cities is much cheaper thanks to Chinese open AI models and Israeli smart tools, because it rarely needs to punish anyone. Mass surveillance technologies — cameras, phone software, local internet networks — do not have to arrest anyone. They only need the citizens to know they are there. The awareness of being watched stops most action before it begins. The target list Every movement for change on the continent has begun the same way. A group of people decides the risk of being seen in public is lower than the risk of staying silent. They gather. They post. The images travel. Sometimes the government falls. Sometimes it holds. The calculation in both cases turns on a single moment, when strangers become visible to each other and to the world as a movement. The upgrade changes that calculation at the source. AI-enabled surveillance is not a neutral grid cast over a city. It is configured around a target list, and that target list is political. The IDS mapping shows cameras clustered where opposition parties organise, not where ordinary crime is highest. The effect is pre-emptive; it suppresses dissent before it forms. Facial recognition at a bus stop does not need to arrest the organiser; it only needs to make clear that organising is no longer anonymous, and that being identified carries a cost. For instance, a protest planned for Sunday and identified on Saturday — through metadata, social media posts, network analysis and face-matching — rarely needs to be broken up by police. The organisers already know they have been seen, and they know what being seen can cost: arrest, charges filed later in court, or even being banned from government jobs. Often, the protest simply does not form. The logic compresses into a loop: the algorithm flags a person as a likely organiser, and that flag is treated as proof that the event would have occurred; thus, the person cannot be innocent because the alarm itself is the evidence. In a country with thin courts and a politicised security service, this becomes a machine that treats intention as guilt. Digital footprints widen the target list. Through the traces users leave online and through state-controlled communications, the state can build a profile of every citizen, then query it for signs of disloyalty. Not for what the person has done, but for what they have liked, who they have called or messaged, which rally their phone was near, which anti-government hashtag they shared. The quietest and most uncomfortable finding within the Atlantic Council’s report concerns this integration of administrative databases into something broader, a loyalty index, built out of records that were each introduced, separately, for a different reason. The consequence is that reform dies early. CIPESA (the Collaboration on International ICT Policy for East and Southern Africa) documents the chilling effect across fourteen countries; a measurable retreat of expression, assembly, and independent media in places where the surveillance architecture has been built out without rights-based safeguards. The citizen who would have marched, posted, joined the committee, filed the complaint, now calculates and stays home. From North to South Africa, a new kind of everyday atmosphere settles in. The seed of change is aborted, not in a cell but in a bedroom, by a person who has understood that the database remembers, and that the algorithm does not distinguish between the opinion expressed and the opinion held. This is what the upgrade protects most efficiently. Not the state’s capacity to punish. It’s capacity to make punishment unnecessary. Foucault called this the panopticon: a design in which the inmate, uncertain whether the guard tower is occupied, learns to guard himself. Those who refuse While AI surveillance is becoming a powerful tool for authoritarian regimes, there are small counter-efforts, underfunded, and racing against a procurement cycle that has already delivered the cameras and wired the databases. Several journalists and organisations are documenting abuses, litigating test cases, and pushing for rights-based AI governance. Civic technologists are turning the same tools back on the state, including fact-checking generative content, monitoring hate speech, observing elections, and offering open, safe, smart tools for journalists and activists. However, the symmetry is false. The infrastructure, the compute, the training data, the vendor contracts, the technical talent, all sit with the state and its foreign suppliers. They do not sit with the African citizen, who struggles every day for just their daily bread. A civic-tech app might monitor some corruption, but it cannot match the control room. The few cubic centimetres Orwell’s dictatorship needed a ministry, a war, an informant behind every door. The new African version seems to need only a loan agreement for buying AI infrastructure, and a population that has been persuaded that the whole thing is for their convenience. The historian Yuval Noah Harari has warned that the combination of biometric sensors, facial and voice recognition “make it possible for the first time in history for a dictatorial government to follow all the citizens all the time,” and could produce totalitarian regimes “much, much worse than anything we saw in the 20th century.” That warning is usually aimed at Beijing. But Beijing and Tel Aviv are starting to export the tools to Africa. The fear on the continent is not that AI will turn any single African state into a dictatorship. The fear is that AI is lowering the cost of authoritarian capacity and raising its ceiling — at an unprecedented pace, at precisely the moment when the appetite of corrupt regimes is growing, and the legal and institutional safeguards meant to contain them are thin or absent. Certainly, AI can be a lever for development and well-being in Africa; for public services, for innovation, and for expression, too. But, without the underlying infrastructure of democracy and a free press, it becomes a nightmare. Khalid Bencherif is a freelance, award-winning Journalist from Morocco, based in Berlin, specializing in covering environmental and political issues in North Africa. He received the 2022 Michael Elliott Award for Excellence in African Storytelling, given by the International Center for Journalists (ICFJ). This piece was made possible by support from the International Center for Journalists’ Michael Elliott Award , which is celebrating its ten-year anniversary. Written by Guest Contributor View original post (English)