Ethics and safety·September 16, 2026, 02:00

Anthropic: Russian actor used Claude for propaganda in Central Africa

AI-generated and checked against the sources listed below.

Anthropic writes in a new report that a Russian-speaking actor in Bangui used the AI model Claude to produce pro-Russian radio content daily. The account has been shut down.

The AI company Anthropic has published a report on misuse of its AI model Claude. Among other things, the report describes an influence campaign in the Central African Republic (CAR), where a Russian-speaking actor used AI to produce propaganda.

The report covers activities that Anthropic discovered and stopped between December 2025 and August 2026. It came out in September 2026.

Radio station in Bangui

According to Anthropic, a Russian-speaking person in the capital Bangui ran a daily content operation for the radio station Radio Lengo Songo, which broadcasts on 98.9 FM. The person was at the same time working with the Russian state media outlets RT, Sputnik Afrique and TASS.

Claude was used to write stories with pro-Russian, anti-French and pro-government messages, as well as positive narratives about the Russian Wagner Group. According to media accounts of the report, the AI was also used to forge local documents from the Ministry of Defense.

Sources discussing the report link the radio station to Politology. It is an influence unit under the Africa Corps and the Wagner network, which according to them came under the Russian foreign intelligence service SVR at the end of 2023. These are claims by Anthropic and the media citing it, and I have not seen independent confirmation.

More than Africa

The report also covers other Russian activities. Russian state media reportedly used Claude as a kind of automatic editorial assistant to produce tailored content aimed at Moldova, Latin America and Africa, among others. Anthropic also mentions cyber espionage against Ukrainian and European authorities as well as development of targeting software for drones.

What does it mean?

It shows that ordinary AI tools can make propaganda cheaper and faster to produce, including in local languages and in small local communities where radio is still an important source of news. It can become harder to tell what is genuine news.

Anthropic says the company has shut down the accounts that were found, used the lessons to strengthen its safety systems and shared information with authorities and industry partners.