|
Getting your Trinity Audio player ready...
|
Deepfakes, Hidden Prompts and the 2027 Vote: What NITDA Is Telling Nigerians About AI Risk
Nigeria’s technology regulator has spent much of this year delivering the same message in different venues: artificial intelligence is spreading through the economy faster than the safeguards around it, and ordinary users, companies and political actors are going to pay for that gap unless they change their habits. Taken together, the warnings from the National Information Technology Development Agency form a fairly clear picture of where officials think the danger lies.
The most recent signal came in September, when NITDA Director General Kashifu Inuwa endorsed the creation of a centralised National Crisis Communication Hub to counter fake news, hate speech and AI-driven misinformation. He cautioned that AI-generated deepfakes and automated propaganda are likely to intensify before and during the 2027 political season, which is why he argued a coordinated national response is more urgent than ever. His reasoning is a simple one about human behavior: there is a direct correlation between novelty and virality, meaning the stranger and more shocking a fake clip looks, the faster it travels. NITDA also proposed regional symposiums to widen grassroots participation, plans to talk with global technology companies about faster categorisation and takedown of content that threatens national security, and closer work with the cyber units of critical institutions to build layered defenses.
Earlier in the summer, Inuwa told an audience that attacks now run in two directions: they target AI systems themselves and they use AI as a weapon. He singled out advanced social engineering, especially deepfakes used to impersonate executives, officials or relatives, and stressed that cybersecurity depends on collective responsibility, since a weakness inside one organisation can expose an entire network. That theme had already featured at the GITEX Africa summit in Morocco in the spring, where he said AI-powered digital threats are becoming more elusive and destructive and described Nigeria’s move toward what he called total resilience, a defense that involves every level of society and not just IT departments. He called cybersecurity a strategic imperative for national development and pointed to data suggesting that roughly 95 percent of digital breaches trace back to simple human mistakes, which is why he argued that the best encryption is worthless if the user is compromised. Officials also described tools such as dark web monitoring and AI-enabled threat detection as part of the response, alongside the 3 Million Tech Talent program meant to train Nigerians in cybersecurity, data science and AI.
Not all of NITDA’s cautions are about criminals. In June, Inuwa warned during a visit by the cloud company Layer3 that Nigeria risks being left behind in the AI revolution if it doesn’t act strategically, and he tied that to bias. His point was that AI itself is not biased, the data fed into it is, so countries that are poorly represented in training data could end up with systems that serve them badly. The remark sits alongside growing global worry about digital sovereignty and the concentration of AI infrastructure in a few countries and multinational firms, and it explains why he keeps pushing partnerships with indigenous technology companies.
The agency has also issued blunt consumer-facing guidance. Late last year it published an advisory about weaknesses in ChatGPT that could let malicious instructions hidden inside seemingly harmless web pages trick an AI model into leaking user data, a technique known as prompt injection. The agency’s statement put the lesson plainly: an AI interface is not a private vault. Its recommendations were concrete. Users should avoid entering personal information, passwords, financial details or trade secrets into any public generative AI tool, and they should be cautious when asking a chatbot to summarize content from untrusted websites, because those pages may contain hidden commands.
Read together, the warnings point to a few practical steps. For individuals, treat any urgent audio or video message, especially one involving money or politics, as unverified until confirmed through a second channel, because a convincing voice or face is no longer proof of identity. Keep sensitive details out of chatbots. For organisations, train staff, since human error remains the most common way in, and assume that attackers will use AI to make phishing messages more convincing. For political parties, media houses and platforms, prepare verification and rapid-response systems before campaign season, not after a fake goes viral.
There is also a policy dimension. Inuwa has argued that governments need ethics, codes and standards for any system that makes decisions on behalf of the state, and Nigeria has already adopted a Code of Practice for interactive computer service platforms and internet intermediaries aimed at making the online space safer, including for children. How much weight those rules carry will depend on enforcement and on whether the agency can secure cooperation from the global platforms where most of this content spreads.
What stands out is how consistent the message has been. Whether the audience is a summit in Marrakech, a cloud company in Abuja or users of a popular chatbot, NITDA keeps returning to three ideas: AI risk is mostly a human and organisational problem, the political stakes will rise sharply as the 2027 election approaches, and Nigeria cannot afford to treat safety and participation as separate questions. The next few months will show whether the planned crisis hub and platform agreements arrive in time to matter.
