Exposed: How AI was used to manufacture online support for CS Wandayi

US technology firm Anthropic, which owns the popular Claude AI platform, has released a report detailing a domestic political influence operation in Kenya that used its technology to manufacture political messaging praising Energy and Petroleum Cabinet Secretary Opiyo Wandayi.

The report also reveals that AI was separately used to produce content pushing messaging that the Opposition was fracturing ahead of the 2027 General Election, bringing into focus the role AI will play in shaping political messaging in Kenya.

Anthropic said it had identified and removed the account used by a single actor to mass-produce Kenyan political content. The operation, it said, was specifically designed to conceal the coordination behind the messages, with the actor instructing Claude to generate batches of exactly 50 tweets that would read like ordinary comments from Kenyans rather than material from an organised political campaign.

A significant portion of the AI-generated content praised Mr Wandayi over his role in stopping a scheduled Kenya Power electricity tariff increase. The posts used the hashtags #PowerReliefKE and #PoweringTheNewKenya in an apparent effort to boost their visibility and create the impression of widespread public approval for the Cabinet Secretary. Anthropic also found that inauthentic accounts were used to amplify Wandayi-related content.

"Our investigation revealed a highly structured political astroturfing effort that pushed identical, unified messages regarding the tariff increase across different conversations. We evaluated the impact of the network using the Breakout Scale and classified it as Category One. The activity was completely isolated within the network of fake accounts and local influencers on a single platform, failing to reach or influence any real people," the report says.

Nation reached out to Mr Wandayi seeking his response to Anthropic's findings, including whether he or his office was aware of, commissioned, funded or coordinated the AI-generated campaign praising him over the Kenya Power tariff issue. We also asked whether he or his associates had any links to the accounts involved, and whether he had information on the actors behind the accompanying narrative portraying the United Opposition as divided.

CS Wandayi had not responded to the questions by press time. 

Gachagua, Uhuru targeted

But the operation was not limited to promoting the government side of the political debate. Anthropic said the same network pushed claims that the United Opposition coalition was breaking apart ahead of the 2027 election, targeting former Deputy President Rigathi Gachagua and retired President Uhuru Kenyatta.

"The network pushed stories claiming that Kenya's United Opposition coalition was breaking apart before the 2027 general election, directly targeting politicians Rigathi Gachagua and former President Uhuru Kenyatta," the report reads.

The company described the operation as using a single playbook to amplify both pro-government and anti-opposition narratives, a combination it said was consistent with a coordinated electoral communications effort.

The findings provide an early glimpse into how artificial intelligence could change political communications ahead of 2027. Rather than simply generating political advertisements or speeches, AI was being used to manufacture the appearance of public opinion itself, turning predetermined political talking points into large volumes of posts designed to look as though they had originated independently from ordinary citizens.

The actor supplied Claude with the topics, talking points and hashtags, after which the model converted them into 50 themed, character-counted posts ready for publication across platforms. In some sessions, the posts were packaged into an interactive "copy-all" tool, allowing them to be easily deployed. Anthropic says the technology's main value to the operation was therefore scale and authenticity — the political messages themselves were decided by the human operators before Claude was brought into the process.

The report also flags the growing use of fake journalists and political spokespeople, with AI being used to generate profile photographs, invent biographies and create fictional personas designed to appear credible. Anthropic says it also found cases where real people and institutions were impersonated.

The company also found that the same workflow was being used for Kenyan retail brands under the marketing personas "SHANKI" and "Elkins Marketer". A promotional broadcast was repackaged as apparently organic tweets, with links inserted at regular intervals. Anthropic said this resembled a pattern it has observed in Kenya, in which agencies pay local influencers to manipulate narratives.

Anthropic classified the Kenyan operation as Category One on its Breakout Scale — the lowest tier, reserved for activity that stays confined to a single platform and fails to reach or influence real audiences.

Anthropic says AI is increasingly being inserted into influence operations as a kind of newsdesk or content factory, allowing relatively low-resourced actors to produce material at a scale that would previously have required a much larger operation. It has also observed actors using AI to create fake personas, strip away signs of automation and make commissioned or state-linked narratives appear to come from independent voices.

The Anthropic findings come amid a Kenyan digital environment where political narratives are already being produced and amplified outside traditional media institutions.

KICTANet report: Opposition was targeted

A KICTANet report on Kenya's information ecosystem found that alternative platforms such as TikTok, Facebook and YouTube have become central to political communication, with influencers and digital commentators increasingly shaping how audiences interpret political developments.

The report documented content portraying divisions within the opposition, including commentary suggesting that disorganisation within ODM and the wider opposition would hand President William Ruto an easy victory in 2027. One TikTok commentary, for example, framed opposition leaders as pursuing their own interests and capitalising on internal divisions.

KICTANet also documented fabricated media content around the opposition, including a fake article attributed to The Standard newspaper claiming there was a power struggle between Mr Gachagua and the Uhuru Kenyatta camp over the opposition's 2027 strategy. The report found that the fabricated article was designed to borrow the credibility of an established media house and make an unverified political narrative appear legitimate.

"The content exploits the visual identity and authority of a well-known media house to create a false sense of credibility. Users encountering the post are likely to assume authenticity based on branding alone, without verifying the source. Impersonation of mainstream media functions as a credibility shortcut that allows misleading narratives to gain rapid acceptance," the report says.

The wider concern is that such narratives can gain traction because Kenya's digital audiences are increasingly consuming political information through platforms driven by engagement, emotion and personality.

KICTANet says repeated exposure to politically aligned content can reinforce echo chambers and deepen political divisions, while the combination of coordinated amplification and fragmented trust creates conditions in which political narratives can be strategically shaped to influence public perception.

"The use of a full front-page layout, rather than a standalone quote or caption, significantly enhances the perceived legitimacy of the content. Visual replication of an established newspaper format transforms a fabricated claim into high-impact, persuasive political content," reads an analysis from the report.

Follow our WhatsApp channel for breaking news updates and more stories like this.