In May 2026, Nigeria received an uncomfortable preview of what its next general election could look like. A recording that appeared to feature President Bola Tinubu circulated widely on social media. In it, a voice resembling the president made politically sensitive remarks about opposition parties, insecurity, public borrowing and the 2027 polls. It was fake, and the confusion it caused revealed how poorly prepared institutions are for AI-generated deception, according to a detailed Premium Times investigation published on October 10.
How the episode unfolded
The presidency condemned the recording quickly. Presidential adviser Bayo Onanuga called for the prosecution of social media commentator Martins Otse, popularly known as VeryDarkMan, whom he accused of spreading it. The very next day, the fact-checking organisation FactCheckHub established that Otse had not posted the manipulated recording at all.
Someone had taken footage from one of his Instagram videos, layered AI-generated audio mimicking the president's voice on top, and released the altered clip. Roughly a week later, police arrested a man named Ifechukwu Dennis, whom the presidency identified as the originator. By then, though, the controversy had already gripped national attention, and the government had publicly sought the prosecution of someone who was not the source.
The presidency was at once the injured party, the accuser and the first responder, and it named a culprit before anyone had established where the audio came from.
The real problem is speed
Nigeria has lived with propaganda and fabricated claims for decades. What artificial intelligence changes, the Premium Times analysis argues, is how cheaply convincing material can be produced and how far it can travel. Cloning a voice now takes minutes and a few seconds of sample audio. Video manipulation is becoming similarly accessible.
The gap between creation and verification is where damage happens. Convincing synthetic content can be produced far faster than it can be authenticated. In that window, institutions act on what they have not verified, journalists report claims of unclear origin, and the public forms opinions based on fabricated evidence. Corrections, when they arrive, rarely travel as far as the original falsehood.
Why 2027 raises the stakes
Campaigning for the 2027 general election is already intensifying. Opposition figures are setting out platforms; Omoyele Sowore of the African Action Congress, for example, told AIT this week that his administration would introduce no new taxes and would instead recover unpaid oil royalties he put at $62 billion. As rhetoric heats up, the incentive to fabricate statements, scandals or concession speeches will rise sharply.
Several scenarios worry analysts. A fake audio clip of a candidate insulting an ethnic or religious group, released days before voting, could inflame tensions in a country with deep regional sensitivities. A fabricated video showing ballot stuffing could undermine trust in genuine results. A counterfeit announcement from the Independent National Electoral Commission postponing voting in certain states could suppress turnout.
The opposite danger: the liar's dividend
Experts also warn about the reverse effect. As awareness of deepfakes grows, politicians caught in genuine recordings can simply claim the evidence is fake. This so-called liar's dividend erodes accountability, because authentic leaks and investigative journalism lose power when everything can be dismissed as synthetic. Protecting the credibility of real evidence is therefore as important as debunking false material.
What preparedness looks like
The Premium Times piece points to clear lessons from the May episode. Authorities should verify before accusing. Government communication teams need access to forensic tools and partnerships with fact-checkers so they can assess suspicious content quickly. Newsrooms need protocols for handling audio and video of uncertain origin, including clear labelling when authenticity has not been confirmed.
Electoral bodies can help by establishing verified channels for official announcements and communicating them widely before campaigns peak. Platforms such as WhatsApp, X, TikTok and Facebook carry responsibility too, particularly for rapid labelling and reducing the spread of flagged content during sensitive periods. Several countries that held elections in 2024 and 2025 tested approaches such as election integrity war rooms with representatives from platforms, regulators and civil society.
Media literacy may be the most durable defence. Teaching citizens to pause before sharing, to check whether reputable outlets have confirmed a claim, and to recognise common signs of manipulation can blunt the impact of even sophisticated fakes.
Law and its limits
Nigeria has laws covering cybercrime and defamation, but many were written before generative AI became widespread. Lawmakers may consider updates that specifically address non-consensual synthetic media and election-related deception. Any new rules must be carefully drafted, however, because overly broad legislation can be used to silence critics and journalists, a concern civil liberties groups have raised about previous cybercrime prosecutions.
A test for democracy
Nigeria's 2027 election will be one of the first major African polls held in an environment where convincing AI fakes are cheap and abundant. How institutions, media and voters respond will offer lessons for the entire continent. The May episode showed that the technology is already here. The coming months will reveal whether the country's defences can catch up before ballots are cast.