Interest in AI-driven drone autonomy is accelerating across Africa’s commercial drone sector. On the military side, the picture is narrower, messier, and considerably more consequential — and worth separating carefully from the sales pitch.
“Tactical autonomy” has become one of the most overused phrases in defence marketing, and one of the least precisely defined. It gets applied equally to a drone that follows pre-programmed waypoints and avoids obstacles, and to a drone that identifies a target, decides it meets a strike criterion, and engages it without a human confirming that decision in real time. Those are fundamentally different capabilities with fundamentally different consequences, and the distinction matters more in Africa than almost anywhere else, because the continent is simultaneously the site of some of the world’s most-documented recent uses of autonomous and semi-autonomous strike systems, and one of the regions least equipped, legally, institutionally and technically to govern them.
Automation is not autonomy
Industry engineers increasingly draw a hard line here. Pre-programmed flight paths, collision avoidance, return-to-home behavior and waypoint coordination, the features most commonly marketed as “AI-powered” are, in the words of one autonomy-systems executive interviewed by Breaking Defense this year, simply “table stakes,” better described as automation than autonomy. True tactical autonomy, on that stricter definition, means a drone carrying real-time artificial intelligence on board that can interpret what its sensors are seeing and respond to a changing battlefield situation without an operator directing every movement, recognizing a target class, adjusting course around a new obstacle or threat, or coordinating with other drones in a swarm, all without a continuous human instruction for each step.
That distinction is not academic. It is the difference between a surveillance drone flying a pre-set grid over a border post and a loitering munition that hunts, identifies and strikes on its own judgement. Both get called “autonomous” in vendor literature. Only one of them raises the legal and ethical questions that have made autonomous weapons a live subject at the UN.
What has actually been documented
Africa already has real, if limited, history with the sharper end of that spectrum. A March 2021 UN Panel of Experts report on Libya found that forces aligned with the internationally recognized government had used Turkish-made STM Kargu-2 loitering munitions against retreating fighters loyal to Khalifa Haftar in 2020, describing the drones as capable of pursuing targets without requiring a continuous data link between operator and munition. UN investigators stopped short of confirming the drones acted fully autonomously in that engagement, and other analysts have argued the Kargu-2 is better classified as a loitering munition than a true autonomous weapon, but the case remains the most-cited instance anywhere in the world of a drone potentially engaging human targets without direct human authorization of the individual strike.
Sudan’s civil war has since overtaken Libya as the continent’s clearest testing ground. The Armed Conflict Location and Event Data Project has recorded over 1,000 drone strikes in Sudan between April 2023 and early 2026, the large majority flown by the Sudanese Armed Forces using Turkish, Chinese and Iranian-supplied platforms. In September 2024, researchers documented what they described as the war’s first coordinated drone-swarm attack, when a formation of small drones flown as improvised loitering munitions struck Sudanese Armed Forces positions at El Fasher, prompting SAF-aligned units to begin fielding Chinese-made counter-swarm jamming systems in response. Researchers tracking the conflict have separately pointed to strikes such as the December 2024 attack on a hospital in Kalogi, which killed more than 110 people, as evidence of how quickly algorithmically assisted targeting can produce mass-casualty outcomes that oversight mechanisms are not built to keep pace with.
Nigeria, Kenya and South Africa: three different starting points
The three countries most often cited as leading Africa’s AI-drone adoption are, on closer inspection, at three quite different stages — and mostly not in the military domain.
South Africa has gone furthest toward building institutional capacity specifically for military AI. The government has stood up a Defence Artificial Intelligence Research Unit (DAIRU) to pool government, private-sector and military resources for AI-related defence projects, and South African manufacturer Paramount Group has fielded the N-RAVEN, a family of small drones explicitly built around autonomous swarm coordination in units of up to twenty aircraft, marketed around local manufacture and technology transfer to partner militaries rather than sold as a finished import. That model speaks directly to the “no legacy systems to unwind” argument often made for African adoption, since it is designed to be built inside a customer’s own defence-industrial base rather than bolted onto it.
Nigeria’s military services have each announced intentions to bring AI into their operations, including AI-enabled drones, and Nigeria has begun collaborating with Ethiopia on drone development tied to what is billed as Africa’s largest drone manufacturing facility, near Abuja. But by the assessment of researchers who track the continent’s drone programmes closely, Nigeria has not yet fielded a drone with genuine autonomous targeting or navigation capability; its most advanced systems remain human-piloted or remotely operated platforms, however sophisticated their sensors.
Kenya is the outlier of the three, and its absence from the tactical conversation is itself informative. Kenya’s most visible drone-AI work is concentrated almost entirely in civilian applications — precision agriculture, infrastructure and pipeline inspection, and disaster and emergency response — the same categories, notably, that tend to get cited whenever someone wants to argue Africa is moving fast on AI-drone adoption in general. That is a genuine and significant trend. It is just not a military one, and conflating the two risks overstating how far tactical autonomy specifically has actually travelled into African armed forces.
Why the interest is real anyway
None of this means the underlying appeal is manufactured. African militaries have concrete reasons to want autonomy that reduces reliance on a continuous radio link between pilot and aircraft. Electronic warfare and GPS jamming have become standard features of the conflicts in Sudan and the Sahel, and a growing share of commercially available autonomy software, platforms like Applied Intuition’s recently launched Drone Stack and Tycho.AI’s Voyager system, both aimed squarely at defence integrators, are built specifically around GPS-denied navigation and onboard processing that keeps a drone flying and functioning if its communications link is jammed or cut. For militaries already struggling to counter cheap swarming drones and loitering munitions, as Sudan has demonstrated at scale, autonomy that survives jamming is not a luxury feature; it addresses one of the most immediate operational vulnerabilities in the region’s current conflicts. The same resource-constrained environment that limits legacy infrastructure also removes some of the institutional inertia that slows adoption in better-equipped militaries elsewhere — a genuine, if double-edged, advantage.
The regional exercise circuit reflects the same pull. At AFRICOM’s African Lion exercise in Morocco this May, participating forces tested autonomous target systems and tactical UAS specifically aimed at compressing targeting and decision-making timelines, alongside partner nations. Even where African militaries are not yet fielding autonomous strike systems of their own, they are increasingly being brought into training environments built around them.
The governance gap
The more serious problem is not whether the technology arrives, researchers who study this closely are largely convinced it will, but whether the frameworks meant to govern its use arrive with it. Analysts at the Sudan-focused African Centre for the Constructive Resolution of Disputes (ACCORD) have argued that the African Union’s peace and security architecture, designed well before AI-enabled weapons entered African conflicts, is now straining to keep pace with technology capable of producing casualties faster than existing accountability mechanisms can respond to them. Writing for Africa Defense Forum, one security researcher put the risk starkly: without a clear African voice in the international conversation shaping norms around autonomous weapons, the continent risks becoming “a proving ground and a casualty for others’ technological ambitions” rather than a genuine participant in setting the rules.
That risk sits uncomfortably alongside the already-documented pattern of civilian harm from existing, far less autonomous drone operations in Nigeria, the Democratic Republic of Congo and Burkina Faso — harm caused by systems that still had a human pulling the trigger. Layering machine-speed targeting decisions on top of an accountability gap that already exists is, by most serious accounts, a matter of when rather than if.
A more useful question than “how fast”
The framing that tactical autonomy is now simply table stakes, something every serious operator must adopt or fall behind, gets the urgency roughly right and the specifics roughly wrong. What is accelerating fastest across Nigeria, Kenya and South Africa is commercial and civilian AI-drone use, not tactical military autonomy, and the two should not be read as the same curve. What tactical autonomy Africa has actually seen, in Libya, and now more seriously in Sudan, has arrived through active conflict rather than considered procurement, ahead of any regional framework built to govern it. For African militaries genuinely weighing the technology, the more useful question is not how quickly they can adopt autonomy once it becomes standard, but whether they, and the regional institutions meant to oversee them, will have any real say in how it gets used before that happens.