The INEC preliminary voter register for the 2027 elections has been found to contain AI-made photographs, and the commission itself says some doubtful entries came from its own registration officers. That raises a larger issue than how the images were produced: how did dubious records find their way into a system intended to determine who is entitled to vote? The Independent National Electoral Commission (INEC) now faces demands to explain how it checks records, who is allowed into its systems, and who is held responsible for failures, and not merely whether it can detect synthetic images.
The INEC preliminary voter register carries over 103 million voters, compared with roughly 93 million in 2023. INEC says the jump comes from fresh registrations, yet continuing doubts about odd records have strengthened calls for a plainer explanation of how raw entries are vetted, and of who is held to account when systemic irregularities get through.
Clement Nwankwo, who leads the Policy and Legal Advocacy Centre (PLAC), went on ARISE News to question how far the register can be trusted. He cited photographs said to have been produced by AI, along with abrupt and unexplained surges in the number of registered voters.
His remarks highlight a wider challenge facing INEC. Deleting doubtful records from the INEC preliminary voter register is only the beginning, and the body must prove that its whole process for gathering, checking and scrutinising voter information can be relied upon.
Registration officers entered doubtful records on the INEC preliminary voter register
Responding to the criticism on 30 October, Lawrence Bayode, INEC's head of ICT (information and communications technology), said the commission was acting to address the irregularities in the INEC preliminary voter register before the final register is published on 15 December 2026.
Even so, what Bayode said about the INEC preliminary voter register revealed a graver flaw in how the system works: some doubtful entries had been entered by registration officers themselves, who held approved entry rights to the core INEC database.
Bayode admitted that registration officers in a number of states had introduced faults into the register, among them AI-generated images and badly handled registration. These faults now sit in the INEC preliminary voter register.
Speaking about the officers, Bayode stated, “They have access to register people.” He added that the commission detected these inconsistencies through its ABIS, an automated biometric identification tool that compares biometric records without any human involvement.
The remark weakens the notion that weaknesses in voter lists stem mainly from generative AI. Although such tools make synthetic images easy to produce, INEC's own account points elsewhere. The problem lies in the privileges held by insiders, meaning the people allowed to operate the terminals used for registration, and in the governance rules that restrict what they may do.
This lapse raises questions about how the commission supervises its own operations.
- What checks are placed on the officers who enrol voters as they carry out their duties?
- What validation rules apply before entries are saved into the central records' system?
- Why were suspect records not set aside before the preliminary register was made public?
The episode also shows where automated safeguards stop. Biometric algorithms can spot duplicate fingerprints or unusual features. But how well they work depends on whether the submitted data is sound, whether the workflow is well-designed, and whether flagged anomalies are properly pursued.
INEC has not yet said how many photos created by AI or invalid entries it found, how many staff members were involved, or what steps, whether sanctions or court action, have begun. Without those details, the public cannot properly judge how badly the INEC preliminary voter register was compromised, or whether current safeguards are enough to win back trust.
A larger register and a problem of transparency
The Independent National Electoral Commission (INEC) said its Continuous Voter Registration (CVR) exercise, held in the run-up to the 2027 elections, signed up more than 11 million people. Automated biometric screening then weeded out about 1.2 million entries judged invalid, leaving over 9 million new voters to be added to the 93 million on the books in 2023.
According to Bayode, this automated filtering process produced the initial figure of more than 103 million registered voters.
However, the sharp rise in numbers has stirred public unease, mainly because observers and voters keep pointing to obvious mistakes in the rolls that have been published.
The question goes beyond how big the register has grown. It is also about whether ordinary citizens can check for themselves how the data was updated and confirm how INEC went about it. Given the legal requirements on publication and the errors already found, openness to the public matters a great deal.
Did INEC meet the law on publishing the voter register?
Under Section 19(1) of the Electoral Act, INEC must publish the preliminary register on its web portal and also post it at each local government office, along with every ward where enrolment continues. This has to be done no later than 90 days ahead of polling. With the Presidential and National Assembly polls fixed for 16 January 2027, the cut-off for this step passed on 18 October 2026.
Under Section 20, the commission carries a further obligation. Once it has resolved all disputes arising from claims and objections, and removed any names that appear twice, it must publish the definitive voters' register no later than 30 days before election day, which puts the cut-off around the middle of December 2026.
INEC failed to meet the 18 October date for placing the INEC preliminary voter register online. The initial version appeared on its website only on 26 October, which was 81 days ahead of voting.
The body contended that displaying the register at ward-level locations from 9 to 15 October discharged its duty. However, a physical display on its own does not satisfy Section 19(1), which plainly demands that the register be available both in person and online. The delay reduced the time open to the public for scrutiny and prompted fair doubts about whether the law was followed.
The period of public scrutiny has exposed further faults in the system. Bayode admitted that the commission had been sent complaints about voters whose records were absent, mistakes made in clerical entries, and changes of address that had not been carried over.
INEC told the public these matters are being looked into before the register is published in its final form, but dealing with claims one at a time addresses only a portion of the difficulty. Voters also want openness about how investigations are conducted, the standards used for verification, and the audit trails left behind.
The episode points to a basic lesson in running elections digitally. Help from the public in spotting flaws is useful, yet it is no substitute for strong controls inside the institution itself. Equally, outside scrutiny cannot rebuild trust unless people can plainly see how mistakes are corrected and stopped from recurring.
INEC's AI plans must pass a governance test
All this is happening while INEC extends its use of technology and prepares policies to govern AI across its operations.
According to Bayode, the commission for now confines its use of AI to low-risk tasks, namely chatbots that answer questions from the public and the planning of how election materials are moved. He stressed that none of these tools has any bearing on the core results of an election.
The commission is also putting the finishing touches to an official AI governance framework in advance of the 2027 elections, intended to set the limits within which automated tools may operate and to guarantee that people stay in charge of sensitive decisions.
The commission has also created a specialist AI unit within its ICT division, and it has joined regional initiatives that seek to promote the responsible use of AI during elections.
These moves suggest the commission understands the technical dangers. But the controversy over the voter register shows that good governance cannot rest on software controls alone. It must also cover the way human operators use systems, the auditing of what privileges users hold, and the enforcement of accountability once internal safeguards fail.
Bayode promised that the AI governance framework would be made public. Once published, voters, political parties and groups from civic society will be able to assess for themselves how well INEC's operational safeguards hold up.
INEC's fixes will decide whether trust is earned
The commission has promised to remove synthetic images and settle data inconsistencies by 15 December, and how it handles this will be a major measure of how quickly it can respond ahead of 2027.
However, cleaning up the INEC preliminary voter register only deals with the visible problem. If it wants to win back lasting confidence, the commission must explain how improper data got past system checks, name the weaknesses within the institution, and set out the technical measures it has put in place to stop such breaches happening again.
The row that is still unfolding underlines an important lesson about running public systems digitally: moving a process online does not, in itself, make it trustworthy. Software can speed up administration and highlight irregularities. But where control over who gets access is lax and supervision is thin, the same weaknesses the tools were brought in to cure may simply be repeated, and even magnified.
As 2027 approaches, INEC's task goes beyond rolling out advanced software. The commission must show that both the people and the technology guarding Nigeria's elections are safe from tampering, open to scrutiny, and answerable for what they do.
In the end, the credibility of an election depends less on how many fraudulent records are purged than on whether voters can have confidence in the whole chain along which their information is gathered, checked, and protected.
