OpenAI has dismantled a covert influence operation dubbed “No Bell,” tracing it back to likely Russian actors using generative AI to sway public opinion across sub-Saharan Africa. The campaign, detailed in OpenAI’s February 2026 report, relied on a single ChatGPT account to churn out long-form geopolitical articles and social media posts. These pieces consistently praised Russia while attacking the United States, the UK, and Ukraine. One particularly brazen article argued that Angola’s president deserved the Nobel Peace Prize, a move the operator explicitly admitted was designed to antagonize the administration of US President Donald Trump.
The operation’s infrastructure was as fragile as its narratives. Meta identified and banned several Facebook Pages posing as legitimate news outlets in Kenya, Ghana, South Africa, and Angola. Transparency data revealed admin locations in Russia and Ukraine, with one Kenyan page originally named “Farmtown5” before being rebranded as “Kenya Watchtower.” This suggests the operators bought and repurposed existing digital assets rather than building them from scratch. The content pipeline wasn’t limited to text; the user also experimented with Google’s Gemini for image generation, creating banners for these fake news pages to lend them visual credibility.
To mask the AI origin of their work, the operators employed specific technical tricks. They frequently prompted ChatGPT to avoid em-dashes, a common artifact in machine-generated text, and requested a “human journalist” writing style. Despite these efforts, cracks appeared. A Ghanaian news site published an article where removed em-dashes left behind garbled, nonsensical text. The byline for most pieces was “Dr. Manuel Godsin,” a supposed expert from the University of Bergen. Investigations found no record of this person in Norwegian academic databases. The photo associated with the byline actually belonged to a St. Petersburg law student, exposing the entire persona as a digital fabrication.
This case highlights a shifting tactic in information warfare: low-cost, high-volume AI content targeting specific regional grievances. By focusing on local issues like German arms exports or British military court cases, the operators aimed to embed broader anti-Western narratives within credible local discourse. While Meta and OpenAI successfully disrupted this specific node, the ease of creating fake experts and repurposing old social media accounts remains a persistent vulnerability for digital ecosystems in the Global South.
