A recent report from the AI company Anthropic has shone a light on how artificial intelligence tools are being used in political campaigns in Kenya.
The company found that someone operating in the country used its Claude AI models to create large numbers of fake social media posts that looked like they came from ordinary citizens. These posts were designed to appear as natural, grassroots opinions ahead of the 2027 general election.
According to Anthropic, the person behind this activity generated batches of exactly fifty tweets at a time. They gave the AI clear instructions to make the posts sound like spontaneous comments from everyday Kenyans rather than part of any organised effort.
Many of the messages praised Energy Cabinet Secretary Opiyo Wandayi for stopping a planned increase in electricity tariffs by Kenya Power. The posts used hashtags such as #PowerReliefKE and #PoweringTheNewKenya to try and gain more attention.
The same network also spread claims that the United Opposition coalition was falling apart before the next election. These messages specifically targeted politicians including Rigathi Gachagua and former president Uhuru Kenyatta.
Anthropic described the whole effort as a structured political astroturfing campaign, which means it was an attempt to create the false impression of widespread public support.
Interestingly, the same person used an identical method for commercial purposes. Under a marketing persona called “SHANKI” or “Elkins Marketer,” they generated posts promoting Kenyan retail brands. Some of these were presented as ordinary tweets with product links added every fifth post.
Anthropic stressed that it found no evidence linking the activity directly to the government. The company assessed it as a domestic Kenyan operation that appeared aligned with the ruling coalition because of its pro-administration tone and choice of hashtags. However, the exact organisation or individuals responsible have not been identified.
The company classified the impact of this network as low. The fake posts remained mostly within a closed group of inauthentic accounts and did not appear to influence real users on the platform in any meaningful way.
Anthropic removed the account involved after receiving a tip from another AI firm and used the case to improve its detection systems.
This case forms part of a wider global pattern. Anthropic’s September 2026 report detailed several influence operations in different countries where AI was used to create deceptive political content.
