Artificial Intelligence (AI) research firm behind the popular Claude chatbot, Anthropic, has shut down Kenyan AI-driven propaganda campaign that praises government, while attacking opposition.
It says a “single domestic actor” used its chat bot to mass-produce government biased propaganda.
“We identified and removed an account that was used by a single actor to mass-produce Kenyan political content. The operation was built to look like spontaneous, grassroots public sentiment,” Anthropic says.
The company chose to track, under the GTG-54004 technological designation, an “actor” that generated batches of exactly 50 tweets created to mimic not only organic but also grassroots public sentiment.
“Actor used Claude across several sessions to generate batches of exactly 50 tweets, explicitly instructing the model to make posts look like spontaneous grassroots commentary rather than a coordinated campaign,” it says.
According to Anthropic, a Kenyan actor handed political script, talking points and hashtags, instructing the AI bot to humanise and format the posts into deployment-ready and cross-platform interactive widgets.
The findings have since prompted calls to action, with experts saying such reports have become familiar across the AI industry even as companies seek to show how they identify and disrupt attempts to misuse their models.
“Now Anthropic (or Claude) have detected Astroturfing in Kenyan Political campaigning,” says Ben Roberts, former CEO of Liquid Telecoms, and long-time Group Chief Technology and Innovation Officer of Liquid Intelligent Technologies.
In mid-July this year, Mr. Roberts wrote on his blog how he learned a new word, and the fight over Africa’s internet numbers.
The word Roberts made reference to is “astroturfing”, which he describes as a “public square” where anyone can join, anyone can post, and make decisions by “rough” consensus.
“It started, as these things often do, with a nagging feeling that something wasn’t quite right,” Roberts says.
He put his email archives in front of Claude, Anthropic’s AI, and asked it to help analyse what he was seeing, starting with 2026, then every year from 2017 onward. Nearly nine thousand emails.
“I used the largest model available to me, Claude Opus 4.8, and the choice mattered. Models like Opus are well suited to this kind of work because they can hold an enormous amount of text in their ‘context’ at once,” he explains.
It is has become clear that the AI model can weigh an email from 2019 against one from 2026 and actually notice that they rhyme, rather than skimming a page at a time and forgetting what came before.
Actually, it reads for the things a tired human misses across thousands of messages: repeated phrasings, timing patterns, near-duplicate arguments wearing different names.
“Give it a haystack and it is rather good at telling you which pieces of straw are suspiciously identical,” Roberts explains.
He adds that Claude AI can show deep patterns: which accounts appeared in bursts around which votes, which “different” people sent near-identical text, which time zones and mail systems the messages actually came from. But it cannot tell you who was at the keyboard. The headers that would prove authorship are stripped from public archives.
“So I can point to patterns and facts, but I will not point fingers at named individuals as puppet-masters. That is simply where the honest line sits,” Roberts points out.
The arrival of automated messaging have now become order of the day, with Anthropic’s latest report revealing that, “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.”
Automated messaging have become order of the day especially in polical arena with Anthropic report now showing that, “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.”
While in the US, Senator Bernie Sanders is demanding a pause on advanced AI development and a ban on artificial superintelligence, Kenyan political class, and ordinary appear (still) not pertabed.
In Kenya, Anthropic has successfully exposed a domestic AI-driven propaganda campaign, where politicians are not afraid to use bots to manufacture, and produce mass fake social media post.
Is it a fitting conclusion that a new face of AI mis-information has arrived? The jury is out.
Leave a comment