With Nigeria’s 2027 general elections still a few years away, the battle over information is already taking shape—and this time, artificial intelligence is in the spotlight. At a forum in Abuja organized by Yiaga Africa in partnership with INEC, Microsoft, TikTok, the Centre for Journalism Innovation and Development, Dataphyte, the MacArthur Foundation, the International Foundation for Electoral Systems, the German Agency for International Cooperation, and other groups, two of the biggest technology companies in the world made commitments that could shape how Nigerian voters experience the next election cycle. The AI and Democracy Forum was meant to be a serious conversation, not a publicity event, and for the most part it lived up to that promise. Microsoft’s senior director for global elections, Dave Leichtman, told participants that the company is ready to support local efforts to monitor election-related threats in real time, including assigning a representative to a locally run election situation room if invited. That willingness matters, because election situation rooms are where civil society and observers track disinformation, hate speech, and suspicious online activity as votes are cast and counted. Leichtman also stressed that Microsoft would treat INEC as the authoritative source for electoral information, using the commission’s final candidate list to restrict AI image generation involving those candidates and to prevent the unauthorized use of their likenesses. He reminded the audience that Microsoft and other tech companies signed the Tech Accord to Combat Deceptive Use of AI in Elections in 2024, and even though that accord formally expired at the end of that year, the underlying commitments continue. All of this comes against a backdrop of serious concern: election stakeholders, civil society organizations, and researchers have been asking whether platforms can actually detect deceptive AI-generated content, moderate harmful material in Nigerian languages, and respond quickly enough to election-related threats. The forum was an attempt to get answers, and while the promises were encouraging, the conversations made clear that the road to 2027 will require much more than press releases. It will require ongoing pressure, transparency, and collaboration.
One of the most honest moments came from Google’s Adewolu Adene, head of government affairs and public policy for West Africa. When asked whether Google’s systems can fully understand harmful content in Nigerian languages, he did not hide behind corporate optimism. He acknowledged that language diversity and a shortage of good training data make it hard for AI systems to recognize abusive expressions in Hausa, Yoruba, Igbo, Pidgin, and the many other languages spoken across the country. Google says it is working with civil society and local stakeholders to improve contextual understanding, but Adene was blunt: “It would be deceitful to stand across you and say that it is perfected and we have the perfect model and solution.” That kind of honesty is important, because the scale of the problem is enormous. YouTube alone receives about 500 hours of video every minute, according to Adene. To deal with that flood, Google relies on a mix of artificial intelligence, human moderators, and user reports. But no system is flawless, and harmful content can slip through—especially when it is hidden inside live broadcasts, local dialects, or seemingly harmless conversations that later turn manipulative. The point is not that technology is useless; it is that we are still learning how to use it responsibly. For ordinary voters, this means you should not assume that everything you see online has been checked or approved. Platforms are trying, but they are not perfect. And the more participants pressed, the clearer it became that perfection is still a long way off. But this is not just a problem for Google or Microsoft; it is a problem for everyone who cares about credible elections. The companies have the resources and the reach, but they cannot solve this alone. They need local knowledge, independent researchers, and willing users to report what they see. The forum made it clear that the relationship between tech giants and the societies they serve is still a work in progress. Adene also disclosed that Google had received recognition from NITDA for its reporting on the implementation of the code of practice, suggesting a level of engagement that goes beyond public statements.
The interactive session gave civil society and researchers a chance to push back, and they raised questions that were both specific and uncomfortable. Iveli Ube, executive director of the Centre for Redefining Alternative Civic Engagement for Africa, pointed out that online gender-based violence and digital abuse often happen in languages other than English. She wanted to know how platforms incorporate local expressions and cultural contexts into moderation systems, especially as political campaigns heat up. Rosemary Joines of Digital Africa Research at Saint Kilab added another layer: people are using live broadcasts and comment sections to spread election misinformation, sometimes starting with harmless language and then switching to manipulative content. How do platforms monitor patterns like that? Do they proactively study the tricks people use to bypass moderation? These are not abstract concerns. In past elections, misinformation has fueled distrust, confusion, and even violence. The responses from the tech executives showed both the limits and the potential of their systems. Leichtman said Microsoft runs threat intelligence and threat analysis teams that track cybersecurity threats and disinformation narratives, including those linked to foreign state actors. Microsoft also works with civil society partners to identify misleading narratives and reduce the chances that harmful falsehoods appear in search results or AI tools. He emphasized again that INEC is the authoritative source for election information: “You need to know when, where and how to vote.” The challenge is making sure that authoritative information reaches people before falsehoods do. That means proactively surfacing INEC’s announcements, clarifying rumors, and removing content that violates policies. But as the questions from the floor made clear, current efforts may not be enough. The platforms know this, and they have not yet found a way to translate goodwill into comprehensive, culturally aware moderation. The executives also made a practical case for user involvement: reports from ordinary people can help platforms identify suspicious patterns and understand how their services are being exploited. In that sense, citizens are not just consumers of election content; they are part of the early warning system.
Another key topic was the role of AI chatbots in shaping voter choices. Imagine asking a chatbot for candidate information and receiving a recommendation that subtly pushes you toward one person. That is a real concern, and Leichtman was explicit about Microsoft’s policy: they will tell you about a candidate if you ask, compare positions on issues, but they will not recommend one candidate over another. Adene echoed that idea, describing AI tools as instruments that should support informed decision-making without taking away the voter’s agency. Technology should help people assess information and understand issues, not decide for them. On the question of what happens when things go wrong, both companies talked about reporting mechanisms. Users can flag content they believe is abusive or misleading, and law enforcement agencies have dedicated channels to alert platforms about suspicious accounts, recurring patterns, and harmful activity. Adene was careful to explain that a report does not automatically mean immediate removal; sometimes the platform removes the content, sometimes it restricts distribution, and sometimes it blocks an entire channel, depending on the circumstances. Leichtman also urged people to report inaccurate responses from AI chatbots, pointing to Microsoft’s Copilot feedback mechanism, including the thumbs-down option, as a way for users to help improve the system. Data protection was another part of the conversation. Adene mentioned that Google operates access controls and permissions for the information it collects, and he encouraged users to understand what permissions they grant when using digital platforms. Both companies also have different policies on political advertising, with some platforms banning it altogether. The bigger takeaway, however, is that technology companies are still figuring out how to balance openness, safety, and the integrity of democratic processes. They cannot do it alone, and they are increasingly asking users and civil society to help them identify problems. That is a significant shift from the early days of social media, when platforms often acted as if they were simply neutral conduits without any responsibility for what happened on their services.
The forum also heard about the deep digital divide that remains in Nigeria, especially in rural areas and conflict zones. Participants from Bauchi and Borno States spoke about citizens who lack basic information about voting, let alone digital technologies and AI. For internally displaced persons, the problem is even worse: insecurity restricts physical access, and low digital literacy makes it harder to verify information or even receive voter education. Leichtman acknowledged that Microsoft’s programs for underserved communities in Nigeria are not as extensive as he would like. The company supports technology skills development and has helped expand broadband access to 50 million people across Africa over the past five years, but he admitted that technology companies alone cannot reach every community. Adene pointed to partnerships with civil society groups, academics, researchers, community leaders, and local media practitioners as the way forward. He gave an example from Kaduna State, where AI learning materials were developed using animation and the Hausa language. That kind of local, culturally relevant education is essential if people are to understand what AI can and cannot do. Another practical issue was Google Maps and polling units. Participants asked whether Google was working with local authorities to make sure citizens can locate their polling units, even in remote areas. Adene explained that the availability and reliability of underlying data, as well as regulatory approvals, determine what Google can publish. In other words, the company will not put out information unless it has the necessary permissions and can verify the data. For many Nigerians, these are not minor concerns. If you cannot find your polling unit, if you do not understand the technology that is being used to spread information, if you are displaced and disconnected—then all the high-level promises from tech companies remain abstract. The session made clear that bridging this gap is not just a matter of connectivity; it is about building trust and making sure the benefits of AI are not reserved for people in cities or those who already have access. The companies have the convening power, but they need local partners who understand the terrain and can translate their tools into something meaningful for communities that have been left behind.
The closing remarks at the forum were a reminder that the word “AI” is often used too loosely, and that can cause unnecessary fear. Leichtman pointed out that AI covers many different technologies—chatbots, image generators, back-office tools, and systems used to process information. Some electoral processes in Nigeria have relied on technologies like fuzzy matching of names and optical character recognition for years, long before “AI” became a buzzword. He warned that if we are simply using optical character recognition on forms, we should call it digitising forms, not “using AI.” Clear communication matters, because vague language can make people anxious or, worse, make them switch off completely. Adene agreed, saying the public debate should not focus exclusively on technology but on the people using it and the responsibilities that come with it. The same tool can be used to build or to destroy; it depends on human intention. His message to voters, election officials, and law enforcement agencies was to keep human judgement at the centre—use AI as a support, not a substitute for critical thinking. That sentiment captured the mood of the forum. There is no perfect algorithm that will guarantee clean elections, and there is no single company that can protect democracy on its own. What matters is that governments, tech companies, civil society, and ordinary voters work together, ask hard questions, and hold each other accountable. The promises from Microsoft and Google are a starting point, but the real test will come in the months and years leading up to 2027, when the pressure of a real election, the creativity of those spreading disinformation, and the patience of voters will all collide. If the conversations in Abuja are any indication, at least some of the people building these platforms are willing to listen. The question is whether the rest of us—voters, journalists, election officials, and community leaders—are ready to hold them to their word.