Russian Propaganda Network Employs ChatGPT for African Influence Operations

In a significant development at the intersection of artificial intelligence and geopolitical influence operations, OpenAI has taken action against a network of ChatGPT accounts linked to Rybar, a prominent pro-Kremlin media outlet. The accounts were reportedly using AI tools to generate social media content and develop proposals for covert influence campaigns targeting African nations.

According to a threat report published by OpenAI last week, the campaign—dubbed “Fish Food”—was essentially operating as a “content farm” that produced large volumes of multilingual material later distributed across Telegram and X (formerly Twitter). The company’s investigation determined that at least some of the banned accounts likely originated from Russia.

“In essence, the ChatGPT activity seemed to serve as a content farm for these accounts,” OpenAI researchers stated in their report, while acknowledging they couldn’t independently verify how the AI-generated content ultimately appeared online.

The operation primarily utilized Russian-language prompts to generate content in multiple languages, including English and Spanish. Some of these prompts were specifically designed to create batches of English-language comments that would later be posted by various Telegram and X accounts with no declared association with Rybar, whose name translates to “fisherman” in Russian.

Beyond simple content generation, the operation had more ambitious aims. OpenAI researchers discovered that users had requested help developing commercial plans for covert interference campaigns across Africa on Rybar’s behalf. These plans included managing social media accounts, launching a bilingual investigative journalism website focused on African issues, and arranging paid content placements in French-language media outlets.

One particularly concerning prompt involved editing a proposal for what appeared to be an election interference team. The plan included strategies for building local networks and organizing large-scale events alongside online influence efforts. Other prompts sought information about electoral processes in Burundi and Cameroon, while also discussing campaign options in Madagascar—including suggestions aimed at inflaming protests. According to the report, the most ambitious project outlined in these prompts envisioned an annual budget of up to $600,000.

“The content generated by this operation was typical of covert Russian influence operations over the years,” the researchers noted. “It typically praised Russia and its allies, such as Belarus, criticized Ukraine, and accused Western countries of foreign interference.”

Rybar maintains a significant online presence, with its main Telegram channel boasting approximately 1.4 million subscribers. Despite these sizable audiences, OpenAI indicated they found no evidence that the network’s content had been significantly amplified by mainstream media outlets or that any matching on-the-ground activities had materialized in Africa.

Founded by former Russian Defense Ministry press officer Mikhail Zvinchuk and his associate Denis Shchukin, Rybar has established itself as an influential pro-war military blog. Russian independent media outlets have reported that the late Yevgeny Prigozhin, founder of the Wagner private military company who died in a plane crash in August 2023, was allegedly involved in financing the project.

This case highlights the evolving nature of information operations in the digital age, where AI tools can be leveraged to scale disinformation and influence campaigns. It also underscores the growing focus on Africa as a theater for geopolitical competition, where various powers seek to shape narratives and potentially influence political outcomes.

The discovery comes amid heightened concerns about AI’s role in amplifying disinformation and its potential to be weaponized for political purposes, particularly as numerous countries across Africa prepare for elections in the coming years. OpenAI’s actions represent one of the first publicly disclosed instances of an AI company taking direct action against accounts linked to state-affiliated influence operations.