Anthropic Report Reveals AI-Generated Political Disinformation Campaign in Kenya Ahead of 2027 Elections

AI safety company Anthropic has identified a domestic influence operation in Kenya that used its Claude large language models to generate fake grassroots social media content. The campaign aimed to manipulate public opinion ahead of the 2027 general election by producing numerous tweets praising a government official and spreading narratives critical of the opposition.

The report, "Detecting and Countering Misuse of AI: September 2026," details how a single actor used Claude to create batches of 50 tweets at a time, specifically instructing the AI to make the posts appear organic. These tweets promoted Energy Cabinet Secretary Opiyo Wandayi and propagated claims of disunity within Kenya's United Opposition coalition.

While Anthropic found no evidence of direct government involvement, it assessed the activity as consistent with a pro-administration domestic operation. The company classified the campaign as Category One on its Breakout Scale, indicating limited reach and influence on real people, as the activity remained confined to the network of fake accounts. This incident highlights the growing concern among cybersecurity researchers in Africa regarding the use of AI for election-related information manipulation.

Beyond Kenya, the report also touched on other African cases, including a Russian-linked propaganda operation in the Central African Republic and a commercial "influence-as-a-service" campaign affecting the Democratic Republic of Congo. Anthropic emphasizes its ability to disrupt such operations at an early stage by monitoring content generation on its platform before it reaches social media networks.