On September 24, 2026, INEC Chairman Prof. Joash Amupitan, represented by Enugu Resident Electoral Commissioner Dr. Chukwuemeka Chukwu, told the 22nd All Nigeria Editors Conference in Enugu that generative AI, voice cloning and deepfake video have "democratised mass deception." He asked newsrooms to set up fact-checking desks, use digital verification and forensic tools and verify results only through INEC, according to Voice of Nigeria's report. The warning covers fake videos of electoral officials, cloned audio and doctored result sheets. Nigeria's next general election is in 2027, and the law has not caught up with these tools.
The case for acting now
The argument for fast legislation is strong. A convincing fake of an electoral official announcing a result can spread within minutes of polls closing. Correcting it afterwards may come too late to stop unrest, and voters who see fakes may stop trusting real results. Media stakeholders raised the same alarm days earlier. At the Bagauda Kaltho Media Lecture on September 19, 2026, Premium Times publisher Dapo Olorunyomi warned that AI-fabricated election results and manipulated audio are increasingly common. Regulators who want a legal tool for the worst cases have a fair point.
What the current law does and doesn't do
The Cybercrimes Act, as amended in 2024, contains no definition of synthetic media. Nigeria also has no binding national AI governance law. The federal communications ministry runs a National AI Strategy initiative that stresses ethics, transparency and a human-centred approach. That page does not mention deepfakes, and a strategy document is not enforceable.
One provision does bear on the problem: section 24. It is the cyberstalking clause, and it has a troubled history. On March 25, 2022, the ECOWAS Court of Justice, in a case brought by SERAP, ordered Nigeria to amend section 24 of the 2015 Act. The court found it conflicted with the African Charter and the ICCPR. Rights groups had documented its use against journalists, bloggers and activists. The 2024 amendment removed the "grossly offensive" and "annoyance" language. As the Policy and Legal Advocacy Centre notes, it now covers messages that are pornographic, or that the sender knows to be false and sends to cause a breakdown of law and order or a threat to life.
That narrower wording could plausibly reach a deepfake that incites violence. It leaves other harmful fakes uncovered, and it could still be stretched against critics.
Why a broad "fake content" offence is the wrong fix
The tempting response is a new offence for "false AI content" with wide wording and heavy penalties. Section 24's history is the best argument against it, for three reasons.
- Vague falsehood offences hit speech first. A test based on "falsity" or "harm to public order" hands prosecutors discretion. In a pre-election period that discretion will be used against opposition figures, satirists and reporters, not only against foreign troll farms.
- Authenticity is hard to prove. Forensic detection tools give probabilistic results. Real footage is already dismissed as "AI-generated" by people it embarrasses. An offence that turns on proving a clip is synthetic creates a defence for the guilty and a trap for honest publishers.
- The harmful act is deception, not the technology. Satire, parody and disclosed AI content are legitimate expression. Any rule aimed at the tool sweeps them in.
A proportionate approach would regulate specific, provable harms:
- Impersonation of election officials and fraud on the voting process. Fabricated INEC result sheets and cloned officials' voices are already forms of forgery and impersonation. A tightly drafted offence covering knowingly false statements about how, when and where to vote, or about official results, needs no general theory of AI.
- Intent and harm requirements. Liability should attach only to intentional deception aimed at the electoral process, with clear exemptions for satire, journalism and good-faith error.
- Judicial oversight of takedowns. Removal of content should follow a court order or a reviewable, time-limited process rather than a regulator's opinion.
- Transparency over prohibition. Labelling requirements for paid political advertising that uses synthetic media are less risky than bans.
INEC's own moves point in a sensible direction
INEC says it has created an Artificial Intelligence Division within its ICT Department. It also says it is deploying automated AI auditing tools in results management under a five-pillar AI Governance Framework with mandatory human auditing, per the Voice of Nigeria report. This is the right sequencing: verify at the source, publish results quickly and transparently, and let the authoritative record outpace the fakes. Fast, machine-readable publication of polling-unit results does more against doctored sheets than any criminal statute, because it gives editors something to check against.
The request to newsrooms is also reasonable. Verification desks are cheap compared with litigation, and journalists are the first line of defence. Editors should have clear, fast channels to INEC so that "verify through INEC" is practical on election night.
What lawmakers should do
Nigeria does need to close the definitional gap, but with restraint. The National Assembly should draft any synthetic-media provision as a narrow amendment tied to electoral fraud and impersonation. It should hold public consultation, and it should test the wording against the ECOWAS Court's 2022 reasoning. A rule that would have survived that judgment is one worth passing. A rule that recreates "annoyance" and "falsity" liability under a new label will end up back in court, and journalists will be the ones in the dock.
The threat INEC describes is real, but so is the record of speech laws being misused. A narrow electoral-integrity offence, fast official results and funded newsroom verification protect voters. A broad deepfake ban would put speech at risk without stopping the forgers.