Russian intelligence operatives have been using ChatGPT and Claude to generate disinformation content across Africa at industrial scale, according to a new investigation that uncovered subscription invoices and coordinated influence networks. The discovery reveals how generative AI tools, despite content policies, can be weaponized by hostile state actors to saturate information spaces with anti-Western narratives across multiple countries simultaneously.

What Evidence Did Investigators Find?

Researchers from INPACT and All Eyes on Wagner obtained two invoices from April and May 2026 showing basic-level subscriptions to both ChatGPT and Claude, each costing just over 20 euros per month. The invoices were billed to an apartment in Madrid's tourist district and linked to email addresses associated with Artur "Mirzoian" Tevosyan, identified as a coordinator for Politology in the Central African Republic.

Politology, originally the Wagner Group's influence arm, now operates under direct control of Russia's Foreign Intelligence Service, or SVR. The organization is active in approximately thirty countries and has become the primary vehicle for Russian influence operations across the Global South.

When investigators shared the information with OpenAI, the company confirmed it could link the subscriptions to a cluster of ChatGPT accounts, some now inactive. According to OpenAI, these accounts were primarily used for translation and summarization of press reviews about the Central African Republic, as well as open-source research. OpenAI has since suspended the accounts.

How Are AI Tools Enabling These Influence Campaigns?

The use of generative AI dramatically reduces the operational costs of running disinformation campaigns. Harouna Drabo, an expert on information warfare and journalist, explained the strategic advantage these tools provide.

"While saturating the information space with anti-Western disinformation narratives is the primary operating principle of Russia's influence apparatus in Africa, generative artificial intelligence enables the industrial-scale production of such content. Its use significantly reduces the costs associated with mobilizing the human, technical, and logistical resources required to conduct these operations," stated Harouna Drabo, information warfare expert.

Harouna Drabo, Information Warfare Expert and Journalist

This is not the first time OpenAI has detected such misuse. In February 2026, the company identified a ChatGPT user producing content for multiple African countries, including South Africa, Ghana, Kenya, and Angola. The activity included drafting social media posts for accounts that repeatedly changed identities and generating analytical articles on geopolitical dynamics in sub-Saharan Africa, some attributed to a fictional persona called "Dr. Manuel Godsin".

Anthropic has also identified misuse of Claude for influence operations involving coordination of approximately one hundred social media accounts.

Steps to Understand How These Operations Work

• Content Generation: AI tools translate and summarize existing news about target countries, then generate original analytical articles and social media posts at scale, reducing the need for human writers.
• Identity Obfuscation: Operatives create multiple email addresses and billing identities to mask the true source of accounts, making attribution difficult for platform investigators.
• Coordinated Narratives: Generated content is distributed across dozens of social media accounts and media outlets that appear independent but are actually part of the same influence network.
• Targeting Specific Individuals: AI-generated disinformation campaigns can be weaponized against specific diplomatic staff, as evidenced by smear campaigns targeting French Embassy personnel in Bangui.

What Specific Tactics Were Uncovered?

One particularly revealing detail emerged from the invoices. One ChatGPT subscription was billed under the name "David Denis," not Artur Tevosyan. David Denis was a French Embassy staff member stationed in Bangui from October 2025 to January 2026. Beginning in March 2026, Denis became the target of a coordinated smear campaign in media outlets affiliated with Politology, including Radio Lengo Songo and Ndjoni Songo, as well as regional African media.

The campaign centered on a false claim about a "900 billion CFA" payment allegedly made to the Wagner Group. Using Denis' name on the ChatGPT invoice appears to have been a deliberate choice by Tevosyan, creating a direct link between the AI tool subscription and the subsequent disinformation campaign targeting the diplomat.

A French diplomatic source confirmed to investigators that French Embassy staff routinely face disinformation attacks designed to undermine their work and presence in the country. The Embassy advises staff to seek institutional protection and pursue legal action when appropriate.

What Are the Broader Implications?

The discovery highlights a critical vulnerability in how generative AI platforms are monitored and controlled. While OpenAI and Anthropic have content policies designed to prevent misuse, the low cost of entry, ease of account creation, and difficulty in attribution make these tools attractive to state-sponsored actors. The invoices show that basic-level subscriptions, costing roughly 20 euros per month, provide sufficient access for operatives to conduct large-scale influence operations.

The investigation also demonstrates how AI-generated content can be weaponized not just for broad narrative control, but for targeted attacks against specific individuals. The use of a diplomat's name on an invoice, combined with subsequent disinformation campaigns, suggests a level of coordination and intent that goes beyond automated content generation.

Both OpenAI and Anthropic have taken action in response to the investigation. OpenAI suspended the identified accounts, while Anthropic confirmed it is investigating the matter, though the company has not provided additional details at the time of publication.