Professor Àmúpìtàn: Can INEC successfully deploy AI for the 2027 elections?

By Olarinre Salako,

WHEN the Independent National Electoral Commission (INEC) announced on May 22, 2025, that it had established an Artificial Intelligence (AI) Division to introduce the technology into its electoral activities ahead of the 2027 general elections, I received the news with unease. Not because I opposed the technology.

As a technology leader who has developed and deployed intelligent systems, with patented innovations involving AI applications in vehicle safety and energy optimisation, I recognise AI’s potential to improve efficiency, detect irregularities and strengthen accountability. My concern was the deployment of AI in a sensitive electoral environment without independent validation and robust safeguards. Elections are not ordinary administrative exercises. Failures can undermine democratic legitimacy. INEC already faces a trust deficit following previous electoral controversies. I was unconvinced that INEC could establish and independently validate safeguards within the timeframe.

The voter register: A warning sign

My concerns deepened when INEC announced in October 2026 that Nigeria’s voter register had surpassed 103 million names. The figure was not necessarily fraudulent, but a database of such electoral significance should not be accepted solely on the compiling institution’s assurances. Even in technologically advanced democracies such as the United States, AI deployment in consequential election administration remains cautious and limited. AI influencing voter eligibility, counting or results verification raises distinct questions of accuracy, transparency and accountability.

Following INEC’s publication of its preliminary voter register online on October 26, Nigerians discovered suspicious photographs and entries. Social media user Decosule circulated screenshots of apparently AI-generated photographs attached to voter records in Etsako West Local Government Area of Edo State. The discovery is particularly troubling in Edo State, where Governor Okpebholo has repeatedly promised to deliver 2.5 million votes for President Tinubu in 2027, compared with the 144,471 votes Tinubu received there in 2023. Governors Bago of Niger State and Abiodun of Ogun State have each promised two million votes, exceeding Tinubu’s 2023 performance in their states. APC National Chairman Professor Yilwatda claims that the party has approximately 12.9 million digitally verified members nationwide.

Although these political declarations do not establish electoral wrongdoing, when considered alongside the latest revelations about the voter register, they reinforce the need for independent verification ahead of the 2027 elections. Nigerians deserve to know that every registered voter is genuine and that electoral databases cannot be manipulated by those entrusted with maintaining them.

On October 30, INEC’s Director of ICT, Dr Lawrence Bayode, confirmed on ARISE News that some photographs were AI-generated. More disturbing was his explanation that certain registration officers, using authorised access credentials, had introduced questionable records into the system.

INEC maintains that its Automated Biometric Identification System (ABIS) had already flagged suspicious records and that the register is undergoing verification before certification. Nevertheless, how could artificially generated photographs enter a system supposedly designed to capture and verify real voters? How many questionable records exist? If INEC had already identified these irregularities, why did it not proactively disclose their nature, scale and potential implications before citizens exposed them on social media? The discovery does not establish that INEC’s proposed AI systems generated the photographs or that every affected registration represents a fictitious voter. But it does raise questions about access controls and registration integrity.

Where are INEC’s AI governance safeguards?

During a recent edition of ARISE News’ The Morning Show, Vimbai Mutinhiri recounted her interview with INEC Chairman Professor Joash Amupitan. She said she had questioned him about the commission’s AI governance safeguards, and he responded that INEC was continually refining its processes. However, assurances of continuous improvement cannot substitute for a transparent, independently verifiable governance framework. Responsible governance does not promise perfection; it establishes measurable standards, independent verification, accountability mechanisms and procedures for managing failures.

On October 6, INEC announced that automated AI auditing tools had already been integrated into its results-verification process. Yet its comprehensive AI governance policy was reportedly still being finalised.Why introduce technology capable of influencing sensitive electoral processes before its governance framework is completed and independently scrutinised?

Former INEC National Commissioner Professor Okechukwu Ibeanu has questioned the proposed use of AI to extract election results from physical Form EC8A documents. What happens when electronically extracted figures differ from the polling-unit result sheets? Which record prevails? Who investigates discrepancies? INEC should publish a comprehensive AI governance document identifying every proposed system, its purpose, data sources, operational boundaries, validation procedures and responsible officials. It should explain how the commission prevents unauthorised modifications, maintains tamper-evident audit trails and investigates suspicious outputs.

Independent technical auditors, including appropriately accredited representatives of political parties and civil society, should examine consequential systems under controlled conditions. Such oversight must protect sensitive voter data and cybersecurity.

INEC must establish a Human-in-the-Loop (HITL) framework ensuring that AI supports, rather than determines, consequential electoral decisions. When an algorithm flags discrepancies in election results, who makes the final determination? Can political parties challenge the findings? Will every automated recommendation and subsequent human decision be documented for independent scrutiny?

The controversy also demands forensic investigation. INEC should disclose the number of confirmed AI-generated photographs, affected registration centres, periods of creation and measures taken against responsible personnel. It must also demonstrate that legitimate voters will not be wrongly deleted during database cleansing.

READ MORE AT OLARINRE SALAKO.