Algorithms won’t decide 2027 elections, says INEC chair

Amupitan meant that artificial intelligence may support election work, but it cannot replace voters, electoral officials or democratic institutions. The final choice of who governs Nigeria must come from citizens casting ballots under rules established by law. Technology can assist the process, yet it must remain subordinate to human judgment and accountability.

For example, an algorithm might identify unusual result patterns, help plan delivery routes or suggest information to users. It cannot legally decide which candidate wins, accredit a voter, qualify a candidate or validate results by itself. Those decisions must remain subject to human review, statutory oversight and judicial review.

This principle matters because AI will still shape the election environment. It may affect information, logistics and manipulation tactics. INEC says it has completed 10 of 14 major milestones and is building AI tools, but Amupitan insists no critical electoral decision will be fully automated or delegated to an autonomous system.

An algorithm is a set of instructions for processing information and producing an output. An artificial intelligence system uses algorithms and data to recognise patterns, make predictions or assist with tasks. These systems can work quickly across large amounts of information, but their outputs still require human scrutiny, especially in a democratic process.

In an election, AI could help plan transport routes, manage a biometric database, support voter engagement in several languages or audit election results. It could also detect deepfakes, monitor digital threats and identify unusual patterns. These uses assist officials; they do not give software the legal power to decide an election.

The article presents AI as both useful and risky. It can improve logistics and cybersecurity, but it can also amplify bias, targeted political advertising, synthetic media and automated cyberattacks. INEC therefore says critical decisions will not be fully automated, and its governance framework requires mandatory human oversight.

The 2027 polls will be a vast, multi-level democratic exercise. INEC chairman Amupitan described them as the largest democratic exercise on the African continent. Their scale makes reliable planning, secure information and accountable technology especially important, because failures can affect voters across the country.

INEC expects more than 103 million registered voters to use 176,846 polling units in 774 local government areas. The elections will involve multiple ballots. These will cover the presidency, the National Assembly, governorships and state Houses of Assembly. Each contest adds operational demands for officials, parties and voters.

The commission is preparing for this scale through several initiatives. It has procured additional Bimodal Voter Accreditation System devices, introduced online voter registration and digitised candidate nominations. It has also processed more than 15,000 nominations across 22 political parties, while completing 10 of 14 major operational milestones.

AI influences the environment around an election, not only the formal act of voting. It can determine how quickly information is produced, translated, recommended or circulated. It can also affect transport planning, security and result checking. This matters because voters form opinions within that wider information and logistical environment.

A fabricated audio clip, video or image could be generated and spread to mislead voters. Targeted political advertising could deliver different messages to particular audiences. At the same time, AI might optimise routes or help identify unusual result patterns. The same technology can therefore support election administration or help actors manipulate public understanding.

Amupitan said AI will alter the tactics of democratic actors and anti-democratic disruptors. INEC has created an AI division and integrated automated auditing tools into results verification. It also reports using deepfake detection, digital monitoring and data analytics, while warning that synthetic content could undermine confidence in genuine information.

The proposed safeguard is not to reject AI, but to keep it subordinate to accountable people and institutions. INEC says automated systems may assist electoral work, yet they cannot independently make decisions that determine who votes, which results stand or which candidates qualify. This protects the legal and democratic chain of responsibility.

For instance, an AI tool could flag an unusual result or help verify information. Officials would still examine the finding rather than accepting the machine’s output automatically. Voter accreditation, result collation and candidate qualification will not be fully automated or delegated to an autonomous system. Human officials remain responsible for decisions.

The article identifies three controls: human accountability, statutory oversight and judicial review. INEC’s five-pillar AI governance framework is being developed with mandatory human oversight, while its AI governance policy is being finalised. These safeguards are intended to keep technology answerable to election law and reviewable decisions.

INEC has the central institutional role. It is developing AI governance, operating an AI division and using tools for auditing, monitoring and deepfake detection. Political parties are being urged to adopt an ethical code for AI use. Their conduct matters because campaigns can use digital tools to influence voters and circulate political messages.

The forum brought together media executives, civil society organisations, security agencies and technology companies, alongside electoral officials and party leaders. This shows the intended protection model is collaborative. Media organisations can help scrutinise and communicate electoral information, while civil society groups can examine algorithmic transparency and monitor digital campaigning. Security and technology participants bring relevant operational and technical capabilities.

The article does not assign every group a detailed task. It does state that Amupitan called on parties to adopt ethical rules and urged civil society to scrutinise algorithmic transparency and monitor digital campaign activity. The forum creates a setting for these actors to coordinate before the 2027 polls.

Voters can make a meaningful democratic choice only when the information shaping that choice is reasonably distinguishable from fabrication. If false material is accepted as genuine, citizens may support candidates, reject results or distrust institutions for reasons based on deception. The problem is therefore not merely technical; it affects consent and public confidence.

The article highlights fabricated audio, videos and images as examples. Generative AI can produce synthetic media that appears convincing, while targeted political advertising can deliver tailored messages. These materials may mislead voters before they reach the polling unit. Amupitan described this as “hacking of the voter’s mind,” rather than simply hacking a ballot box.

INEC says its AI unit has deepfake detection software, digital monitoring tools and data analytics expertise. It also has automated auditing tools in results verification. These measures may help track manipulation, but the chairman’s warning makes clear that trustworthy information and human judgment remain essential to informed voting.