Anthropic Discloses Misuse of Artificial Intelligence Model in Pro-Russian African Campaigns
Commercial artificial intelligence systems are increasingly being deployed in state-aligned influence operations to sway public opinion and target political rivals across several African countries.
Suspected Russian operatives used Anthropic's Claude artificial intelligence model to generate propaganda and manage influence operations in the Central African Republic, the United States technology company said in a report. Deutsche Welle reported that the activity was part of broader state-linked efforts targeting African nations.
Anthropic said operators based in Bangui relied on the tool to support the local government and praise Russian mercenaries. The generated content also attacked France and domestic opposition figures. Much of the material supplied Radio Lengo Songo, a local broadcaster funded by Russia, according to investigative collective All Eyes on Wagner.
The report stated that operators asked the system to falsely label opposition leaders as armed rebels. The goal was to provoke local security forces into arresting them. Anthropic said the model refused that prompt, and the company later identified and blocked the accounts.
Lou Osborn of All Eyes on Wagner said Russian networks have run such operations in Bangui for years. Central African authorities previously arrested several figures following online campaigns that accused them of spying for Western intelligence services. Martin Ziguele, leader of the opposition Movement for the Liberation of the Central African People, said smear campaigns against political rivals began expanding in 2020.
Russian representatives dismissed the allegations. Dmitri Sytyi, a Russian mission official in the Central African Republic, said the accusations lacked concrete facts and evidence.
Anthropic documented similar misuse across other African states. In the Democratic Republic of the Congo, operators used the model to craft 318 articles distributed across roughly 70 fraudulent news outlets. More than 250 inauthentic social media accounts then shared the articles to back Congolese officials in their conflict with Rwanda.
The company also found instances of automated content generation in Kenya, Sudan and Mali. In Kenya, the tool created batches of 50 messages designed to support the government while denouncing opposition politicians.
Central African opposition leaders are now calling for a formal inquiry into the foreign artificial intelligence activities uncovered by the company.
