AI and Elections: What Happens When Technology Fails Us?

A voice message comes on election morning. It sounds like an INEC official saying voting has been postponed. A woman in Mbiama, Kontagora, Ife, or Ogoja, who arranged transport and planned to vote, stays home. The message is fake. By the time she finds out the truth, voting has already finished.

In another case, a fake video shows a candidate insulting a community. Anger spreads faster than the correction. Even after people find out it was a lie, the resentment stays.

These examples are just possibilities, but the risk is clear: an election can be messed up without changing a single ballot. Mislead people about when to vote, who to trust, or what happened at their polling unit, and their choices can be manipulated before counting starts.

Nigeria must stop AI from becoming a tool for manipulating elections or an excuse for avoiding responsibility.

This risk made the AI and Democracy Forum in Abuja urgent. YiagaAfrica, INEC, the Centre for Journalism Innovation and Development, and Dataphyte Foundation organized the forum. They asked, “Will Algorithms Determine the 2027 Vote?” For the everyday voter, the question is simpler: will technology protect my choice or help someone steal it?

Some Nigerians might wonder why we should focus on artificial intelligence when issues like intimidation, vote manipulation, vote-buying, security breaches, and impunity already threaten elections. AI is not the root of every election problem. It will not fix misconduct or change institutions that refuse to follow their rules. But it can make familiar issues quicker, wider, and more believable.

The answer is not to panic or be blindly excited. We must ensure that any use of AI that could affect a voter’s rights or an election result is open to scrutiny, challenge, and correction.

INEC’s promise that people, not machines, will have authority over key election decisions is good news. But someone can be in charge on paper while still following whatever a computer suggests. Human oversight matters only when the official knows the facts, checks the results, and can reject them. Honesty from INEC, its officials, and other election-related institutions is essential.

Think about software flagging a voter for being registered twice. It might have found a problem, or it could have made a mistake. Before that person is blocked from voting, someone must look at the evidence and explain the decision. The voter must have a fair chance to challenge it. “The computer said so” cannot settle a question about a person’s rights.

The same rule applies to results. What if technology misreads a polling unit’s result sheet and confuses a three for an eight? Who checks the entry? Which record settles a disagreement between paper and electronic figures? Who fixes the mistake, and when?

These are not just questions for engineers. A wrong number can change the total. An unexplained difference can make a community believe its votes have vanished. Nigerians should not need a degree in computer science to understand how results move from the polling unit to the final declaration, which is currently a big problem in the election chain.

INEC must clearly explain that journey: who records the numbers, who verifies them, who can fix mistakes, and who answers when something goes wrong. Repeated claims that a system works cannot replace proof that it worked in a specific case.

Preparation must go beyond impressive demonstrations. Systems need independent testing. Staff must have practical training. Contractors need to know their responsibilities. Backup plans must work under pressure, not just look good on paper. If technology fails, citizens deserve an explanation and a solution, not a lecture about how complicated it is.

But even reliable election management cannot help a voter who stays home because of a fake announcement.

AI lies can threaten elections in two ways. It can make things that never happened seem real. It can also help politicians deny things that really happened. When faced with a real recording of wrongdoing, a public figure might just say, “It is AI.”

We could end up in a situation where false evidence is believed and real evidence is ignored. People would judge recordings not by their truth but by whether they support their favourite candidate. Eventually, some might stop checking altogether.

That fatigue is dangerous. Democracy suffers when citizens believe lies. It also suffers when they think that truth cannot be found and that participating is useless. Protecting the vote means protecting people’s ability to find out what happened.

The suggested Electoral Information Integrity Coalition could help, as long as it becomes more than just another platform for statements. Election officials, journalists, fact-checkers, civil society, and telecom companies need to have agreements for spotting dangerous lies, verifying claims, and delivering corrections to the affected communities.

Think about a false postponement message going around in a local-language WhatsApp group. An English-language denial hidden on a website will not fix the damage. The correction must reach the same audience through the channels they use: community radio, trusted local voices, shareable audio, and community networks. It must arrive while people can still act.

There must also be a clear line. Criticism of INEC is not automatically false information. Neither is an investigation, an observer’s report, or a citizen’s account of wrongdoing. Protecting voters must never be an excuse to silence them. Those checking claims must show their proof and fix their own errors.

Political parties face a simple test. Condemning a fake video is easy when it attacks your candidate. The harder question is what you do when it hurts your opponent.

Candidates should reject fake impersonation, clearly label campaign audio, images, or videos that are mostly made or changed by AI, and correct lies spread in their favour. Promises against disinformation mean little if campaign teams secretly reward those who spread it. Parties should set consequences for breaches before competition makes dishonesty tempting.

Technology companies must show that their promises work for Nigerians. Can a voter easily report a fake recording? Will the company act before the damage is done? Do its safeguards cover content in Nigerian languages, not just English? A reporting button means little if nobody responds when it really matters.

Local innovators also deserve support, but they should not be exempt from scrutiny. A homegrown product can help observers organize reports or citizens find reliable information. Its origin, however, does not prove its reliability. Nor does an award guarantee election readiness. Testing, privacy assurance, and honest explanations of limitations must come before it is used.

Civil society organizations publishing independent findings should meet the same standard. They should explain their sources, verification methods, and correction processes. Demanding transparency from INEC carries more weight when those making the demand practice it themselves.

For citizens, the responsibility starts with pausing before forwarding. A message does not become true just because it comes from a friend, a religious leader, or a supporter of our preferred candidate. An alarming claim about voting arrangements needs confirmation through established electoral information routes. When we find out that something we shared was false, we should correct it where we spread it.

This does not shift the burden of protection to the voter. Citizens cannot reasonably be expected to investigate every recording while institutions and companies avoid their responsibilities. Public education must support effective safeguards, not replace them.

Before 2027, Nigerians should demand four clear answers. Where will INEC use AI, and who will supervise each use? How can a voter challenge a decision that affects their voting rights? What will parties and platforms do when fake campaign material appears? How will corrections reach affected people quickly, in languages they understand?

These questions must be answered publicly and early enough to expose and fix weaknesses. Election morning is too late to decide who handles a fake postponement notice. A disputed result is the wrong time to find out that nobody knows how to fix a computer-generated error.

Responsibility must not disappear between an electoral body, a contractor, and a social media platform. A citizen seeking help should not be sent from office to office while each organization blames another. Everyone must know who is responsible when something goes wrong.

Used rightly, AI can help voters get reliable information, alert officials to threats, and support evidence examination. Rejecting these possibilities would miss the point. The question is whether technology serves accountable institutions and informed citizens or becomes a shield for those who avoid questions.

We should judge electoral innovation from the voter's perspective. Can a citizen understand the decision? Can the evidence be checked? Can an error be challenged and fixed before the right is lost?

The measure of progress is not how modern an election looks but whether ordinary people can exercise their rights without lies, exclusion, or unanswered questions about results.

Think back to the woman who stayed home because a fake voice sounded official. Her lost chance cannot be fixed by a statement celebrating technological progress. Nor should another citizen lose a vote because nobody would question a machine.

In 2027, no politician, platform, or algorithm should own the truth. Nigerians must remain free to question claims, choose their leaders, and demand proof that their votes were counted correctly. Technology earns its place in democracy only when it strengthens that authority, never when it replaces it.