Propaganda machine: How AI turns political propaganda into new election battleground
National
By
David Odongo
| Sep 16, 2026
Monica Njeri, 73, casts her ballot at Huruma Primary School during the recent Ol Kalou parliamentary by-election. [David Gichuru, Standard]
Kenya’s propaganda machine goes AI: How Ruto’s government is accused of using bots and bloggers to win 2027
A single person with a laptop can now create fake public support for a politician in minutes and at almost no cost. That is the chilling reality Kenya faces ahead of the 2027 General Election, after American AI company Anthropic revealed last week that its Claude model was used to mass-produce fake political tweets praising Energy Cabinet Secretary Opiyo Wandayi.
According to Anthropic, the operator instructed Claude to generate "exactly 50 tweets per batch" that would "read like spontaneous commentary from ordinary Kenyans, not organised political campaign material".
The fake posts praised Wandayi for blocking Kenya Power electricity tariff increases, using hashtags like #PowerReliefKE and #PoweringTheNewKenya. The same network also spread claims that the joint opposition coalition was collapsing, targeting former Deputy President Rigathi Gachagua and former President Uhuru Kenyatta.
Anthropic called it "highly structured political astroturfing fake grassroots support and classified the operation as Category One because it remained confined to fake accounts on a single platform". The company said it found no evidence of government involvement but noted: "the pro-government and anti-opposition messaging was consistent with a single, coordinated electoral communications effort".
Wandayi denied any connection between himself or the Ministry of Energy and Petroleum and the artificial intelligence campaign. He rejected any link to the campaign on Saturday, saying he had only learned of the unverified reports and challenging critics, saying, "What really matters is evidence".
The accusation that President William Ruto's administration is using bots, bloggers and AI to manufacture support is not simply a claim pulled from thin air. It is backed by a documented pattern of state spending, coordinated online attacks and formal government strategy.
The Ministry of Information, Communication and the Digital Economy has budgeted at least Sh100 million annually to pay social media influencers and bloggers to promote government narratives, according to the National Communication Strategy. The plan targets 10 macro influencers who would receive Sh100,000 quarterly and 20 micro influencers who would receive Sh50,000 quarterly, with duties that explicitly include creating hashtags and promoting them on social media.
This is not a new practice. It is simply being formalised. A report by Amnesty International last year revealed that the government had paid influencers to drown protest hashtags and push pro-government messaging, while also intimidating those viewed as dissenters. Amnesty found that a man was paid between Sh25,000 and Sh50,000 per day to amplify government messaging and drown out trending protest hashtags on X.
The propaganda war extends to fabricated content. A fake front page of Taifa Leo newspaper was created and circulated online, claiming Uganda's military chief wanted to buy the Standard Group, a digitally altered image designed to discredit the media house amid tensions with the government.
Sociologist and psychologist Faith Atsango says the problem is not just technology; it is human nature. "Fake news works because it tells people what they already want to believe," she says. "Our brains are wired to accept information that confirms our fears, our hopes and our tribal loyalties. AI makes this worse because it removes the last warning signs. There are no spelling mistakes, no awkward phrasing, no obvious clues that something is wrong. It looks and sounds exactly like the truth."
Atsango warns that politicians understand this better than anyone. "A lie repeated in perfect grammar, from a hundred fake accounts, at exactly the moment a voter is angry or afraid, becomes almost impossible to resist. Politicians do not need to convince you with facts anymore. They just need to flood your phone with the right emotion at the right time. AI has turned propaganda into a science."
She says the danger in 2027 is not that people will believe one big lie, but that they will stop believing anything at all. "When every video can be fake and every voice can be cloned, trust dies. And when trust dies, democracy dies with it."
Lawyer Peter Wena says Kenya's legal framework was written for a world that no longer exists. "The Computer Misuse and Cybercrimes Act of 2018 requires us to prove that someone knowingly spread false information," he says. "But how do you prove knowledge when the person sharing a deepfake genuinely believes it is real? The law assumes a liar. AI gives us something worse: an innocent messenger carrying a dangerous lie."
Wena says the Elections Act of 2011 does not even mention synthetic media. "A radio station can play an AI-generated advert today and there is no law against it. We are bringing a knife to a gunfight. Parliament must define synthetic media, create clear offences and force platforms to act. Without that, 2027 will be a free-for-all."
He adds that the courts are also struggling. "Judges are not trained to understand deepfakes. Prosecutors cannot tell the difference between a cloned voice and a real one. We need specialised training and fast-track courts, or justice will simply arrive too late to matter."
IT expert Pius Maingi says the technology to fight AI already exists but nobody is using it. "We can detect deepfakes. We can trace bot networks. We can flag coordinated inauthentic behaviour in real time," he says. "The problem is not capability. The problem is that the people with the power to act have no incentive to act. The same platforms that spread the lies are the ones making money from the clicks."
Maingi warns that Kenya's election infrastructure is dangerously exposed. "The IEBC has no AI monitoring unit. The Communications Authority is underfunded. The NCIC is reactive, not proactive. We are waiting for a crisis instead of preparing for one. By the time a deepfake goes viral in 2027, it will be too late to contain it."
He says the solution must be public-private. "The government cannot do this alone. Civil society, media houses, universities and tech companies must share data and build a rapid response system. Fact-checking cannot be a hobby anymore. It must be infrastructure."
Kenya has been fighting AI-generated content before. The Oslo Centre, in a report published in May 2025, warned that AI is a double-edged sword that can build and destroy democracy at the same time. The centre said the window to protect elections is closing. The Oslo Centre documented how AI was used to mislead voters in 2022. A deepfake video on TikTok falsely claimed the former electoral commission chair would "open" election servers and that Raila Odinga had won by a large margin. Fake audio clips were attributed to President William Ruto, claiming he admitted his wealth came from corruption. Women candidates, including Martha Karua, were attacked online with deepfakes and gender-based abuse.
A white paper titled Governing the Machine: Artificial Intelligence, Electoral Integrity, and the Crisis of Democratic Trust in East Africa, published by Maxmillian Ochango in April 2026, warns that Kenya's laws cannot handle AI lies. The paper says the cybercrimes law's knowingly requirement has "failed in the face of AI and that the Elections Act does not define "synthetic media." It adds that the IEBC has no AI monitoring unit and key agencies do not share information. "The institutions exist but coordination does not," the paper says. The Oslo Centre wants Parliament to strengthen the cybercrimes law and force political parties to disclose their AI campaign strategies to the IEBC.
Experts like Maingi also fear the "liar's dividend", where real scandals are dismissed as fake news because people no longer know what to believe. Atsango says repeated attacks on the IEBC ⁷ had already damaged trust and AI fakes will make it worse.
A National Council of Churches of Kenya study found that 77 per cent of Kenyans get political news from social media, with WhatsApp the biggest channel. It warned that unverified AI content could erode public trust in electoral institutions and inflame political tensions. The NCCK has been training religious leaders on media literacy with UNESCO since September 2025.
The government launched the Kenya Artificial Intelligence Strategy 2025-2030 in March 2025 but the Oslo Centre noted it does not specifically address elections. In August 2026, the Ministry of Information released a draft AI policy for public comment. A lawsuit by activists was certified urgent by the High Court in February 2026, seeking a full national AI policy to stop election manipulation.
The IEBC has acknowledged the threat. Acting CEO Ruth Kulundu says that "AI deepfakes and other digitally enabled threats can manipulate voter perceptions, incite violence and undermine the credibility of electoral outcomes." The commission is establishing a situation room and a communication desk to counter disinformation.
The Oslo Centre recommends teaching the public, especially the young, how to spot AI fakes, funding real-time fact-checking tools and forcing Meta, Twitter and TikTok to obey Kenyan election laws and quickly remove false AI political content.
"Kenya's Gen Z population is expected to constitute a significant portion of the electorate in 2027," the Oslo Centre said. "This generation is highly active online, making them both targets and key defenders against digital disinformation."
"The question is not whether AI will be used to lie in 2027. It already is. The question is whether we will have the courage to fight back before the votes are cast," says Maingi.