Anthropic Uncovers AI-Powered Political Influence Campaign in Kenya

AI safety company Anthropic has revealed details of a political influence operation in Kenya that leveraged its Claude large language model to generate social media content. The campaign involved a single actor mass-producing pro-government messages and attacking opposition figures, with the aim of making coordinated messaging appear as genuine grassroots sentiment ahead of the 2027 General Election.

The operation used Claude to create batches of 50 tweets, crafted to resemble organic commentary, promoting government initiatives like electricity tariff increases and spreading claims about the opposition's disunity. While Anthropic found no evidence of direct government involvement, the content's alignment with the ruling coalition suggests a domestic Kenyan operation.

Crucially, Anthropic clarified that Claude was used to amplify pre-determined political strategies rather than creating the objectives itself. Operators provided topics, talking points, and hashtags, with the AI model then generating the themed and character-counted posts. This highlights the AI's role in scaling existing communication efforts rather than formulating independent messages.

Beyond politics, the same AI workflow was also adapted for commercial use, generating promotional content for Kenyan retail brands. The discovery, stemming from a tip from OpenAI, underscores the growing challenge for platforms and audiences in distinguishing authentic public opinion from AI-assisted coordinated campaigns, particularly for countries like Kenya facing upcoming elections.