AI Is Making Misinformation More Dangerous — Africa Needs to Prepare

Misinformation did not begin with artificial intelligence. Propaganda, rumours and deliberately fabricated stories have been used for centuries to influence how people think. What AI changes is the economics of deception: one malicious actor can now produce convincing articles, photographs, cloned voices and videos faster and more cheaply than before. The World Economic Forum warns that low-cost multimodal AI has lowered the barrier for malicious actors to produce text, audio and video disinformation at scale.

There is also an important difference between misinformation and disinformation. Misinformation can simply be inaccurate information passed along by someone who believes it is true. Disinformation involves false or misleading information deliberately spread to manipulate or deceive. Jean-Marc Rickli and Tobias Knappe of the Geneva Centre for Security Policy argue that AI is dramatically expanding this threat, particularly as autonomous AI agents become capable of operating with less human supervision.

That distinction matters for Africa because information manipulation can become political infrastructure. False narratives can be deployed to damage candidates, inflame ethnic or religious divisions, suppress confidence in elections or manufacture the appearance of overwhelming public support. AI makes it possible to generate thousands of variations of a message, translate them into local languages and potentially coordinate distribution through automated accounts. The WEF describes experiments in which hundreds of AI agents were able to coordinate autonomously in a simulated social-media environment to promote a political candidate.

The danger goes beyond creating fake content. According to Techview Africa, there is also growing concern about the use of deepfake voices to impersonate public figures for scams and cyberbullying. Rickli and Knappe warn that future influence campaigns could target machines themselves. If AI assistants increasingly become the systems people consult for news, research and decisions, malicious actors have an incentive to contaminate the information those systems learn from. Their analysis points to research showing that as few as 250 poisoned documents could introduce a hidden vulnerability into a language model. The WEF argues that human and machine cognition may therefore need to be treated as critical infrastructure.

Africa should take that threat particularly seriously during elections. A fabricated video appearing hours before voting could falsely show a presidential candidate insulting an ethnic group, accepting a bribe or withdrawing from an election. Even if journalists expose the video as fake several hours later, millions of people may already have watched and shared it. The problem becomes worse because the existence of deepfakes gives politicians another weapon: genuine recordings can also be dismissed as “AI-generated” when they become inconvenient. The WEF notes that deepfakes are increasingly capable of affecting perceptions of political candidates and that merely knowing deepfakes exist can cause people to doubt authentic evidence.

Technology companies are experimenting with different responses. On X, Community Notes allows contributors to attach additional context to potentially misleading posts. A note becomes publicly visible when enough contributors who have historically held different perspectives agree that it is helpful. X has also made the Community Notes algorithm and underlying data publicly available for auditing. X explains how Community Notes works here. It is important, however, not to mistake Community Notes for an AI deepfake detector. People write and rate the notes; algorithms help determine whether sufficient cross-perspective agreement exists to display them. That approach attracted the attention of Meta. In 2025, Mark Zuckerberg announced that Meta would end its US third-party fact-checking programme and move Facebook, Instagram and Threads toward a Community Notes model inspired by X. Meta explicitly said it had “seen this approach work on X”.

Community moderation alone cannot solve Africa's problem. Ordinary users need stronger information literacy. Before sharing an explosive claim, people should ask who originally published it, whether reputable independent organisations are reporting the same thing, whether the account is authentic, whether the photograph or video has appeared previously in another context and whether the post is deliberately trying to trigger anger before giving them time to think. Reverse-image searches, original government documents, reputable news organisations and independent fact-checkers can all help. The WEF recommends asking basic questions such as who is communicating, what they gain from the message, whether the information can be independently verified and what harm could result from sharing it.

African governments also need stronger rules for synthetic political media, disclosure requirements for AI-generated election advertising and rapid mechanisms for reporting impersonation and deepfakes. But governments should not be given unrestricted authority to decide what counts as truth. Such powers can easily become censorship tools against journalists, activists and political opponents.

Nigeria's #EndSARS experience illustrates why independence matters. Following the events at Lekki Toll Gate on October 20, 2020, a Lagos State Judicial Panel later found that soldiers and police shot peaceful protesters; Amnesty International said its own investigation had concluded that security forces killed protesters at Lekki and Alausa. The Lagos State Government subsequently rejected the panel's finding that nine people were killed at Lekki, arguing that the conclusion was inconsistent with the evidence before the panel. The government's position was reported here.

That dispute demonstrates precisely why a misinformation framework should not make governments the sole arbiters of disputed facts. Governments themselves are participants in political communication and have incentives to protect their reputations. The same applies to opposition parties, corporations and influential individuals. Africa therefore needs independent, legally protected information-integrity institutions involving journalists, universities, technologists, civil-society organisations, election bodies and independent fact-checkers. Such institutions should publish evidence and methodologies transparently, allow findings to be challenged and apply the same standards whether a questionable claim originates from an anonymous social-media account, a presidential candidate or the government itself.

AI can also be part of the defence. Detection systems can identify coordinated behaviour, cloned media and suspicious amplification patterns. The WEF suggests defensive AI agents could monitor coordinated disinformation and identify patterns unlikely to originate from humans. But no detector should automatically become the final judge of truth. AI models make mistakes too. The deeper challenge is trust. If Africans reach a point where every embarrassing video is dismissed as a deepfake, every government statement is assumed to be propaganda and every social-media rumour is treated as equally credible as documented evidence, misinformation has already succeeded.

Africa's response should therefore protect two things simultaneously: freedom of expression and the ability to establish facts. Independent oversight, transparent evidence, media literacy, accountable technology platforms and carefully designed AI regulation offer a better path than government-controlled truth ministries.

In the AI era, protecting information may become as important to democracy as protecting ballot boxes.