Anthropic report: Is Russia using AI for disinformation in the Central African Republic and elsewhere?

The Claude app was allegedly misused by Russian agentsImage: Samuel Boivin/NurPhoto/picture allianceAdvertisementIn Bangui, the capital of the Central African Republic, Russian agents are said to have deliberately used the AI assistant Claude to promote pro-Russian influence.  A report titled "Detecting and Countering Misuse of AI," published by US-based company Anthropic in September, detailed how opposition politicians were monitored, propaganda content was created and disinformation campaigns were supported. 

The report claimed that Claude was used primarily to produce content broadcast by Radio Lengo Songo, a radio station that, according to the international investigative collective All Eyes on Wagner, is funded by Russia. The collective, made up of journalists, academics and experts, investigates the activities of the Russian mercenary group Wagner Group and its successor structures, such as the Africa Corps, around the world.

The Anthropic report said the content was pro-Russian and portrayed the Central African government and the Africa Corps mercenaries in a positive light, while taking a critical stance on France and the opposition.

Lou Osborn, a member of All Eyes on Wagner, was involved in the research for the latest report. He believes these methods are part of a practice that has already been observed in connection with the Wagner Group in the Central African Republic. "There have been several arrests of people who were not necessarily closely affiliated with the Touadera regime," he told DW.  "They’ve been doing this for several years now."Faustin-Archange Touadera has been president of the Central African Republic since 2016.

"The arrests were the subject of online campaigns in newspapers and on the radio which portrayed those individuals either as proxies for France or Western intelligence services, or as people preparing a coup against Touadera," Osborn added.

Anthropic also reported that Claude was prompted to portray opposition figures as armed militants in an effort to provoke security forces into taking action against them. Anthropic says Claude refused the request. However, the report did not name the opposition figures who were monitored, specify the exact nature of the data collected, or say who received it.

For the opposition, the Anthropic report underscored the determination of the government in Bangui and its Russian allies to get rid of them. Martin Ziguele, chairman of the Movement for the Liberation of the Central African People, referred to an earlier case. "We have been observing this since 2020, when the CPC, a coalition of rebel groups, attacked Bangui. Strangely enough, several media outlets did everything in their power to establish alleged links between certain opposition leaders and the CPC," Ziguele told DW. "There were also attacks against the opposition claiming that we were being paid by foreign countries. There was an attempt to pass a law stating that foreign agents were present in the country. All of this relates to the activities of the Wagner Group, as does this latest revelation."

DW also sought a statement from Russia. "Yes, I heard about [the report] in the news," said Dmitri Sytyi, a representative of the Russian mission in the Central African Republic. "But there are no concrete facts in it. [...] These are vague allegations, without precision and without evidence."

Russia firmly denies the allegations, but the report indicates that the Central African Republic was not an isolated case. According to the report, the Democratic Republic of the Congo (DRC), Kenya, Sudan and Mali were also affected.

In the DRC, Anthropic identified 318 articles that, according to the report, were published by a network of about 70 fake news sites and subsequently distributed by more than 250 fictitious accounts on X. The content supported the positions of the Congolese government against the backdrop of the conflict with Rwanda.

In Kenya, Claude was allegedly used to generate batches of 50 messages at a time that supported the government and were hostile toward the opposition.

In Sudan, Anthropic says AI was used as part of an operation attributed to individuals in the United Arab Emirates — particularly to create fabricated witness statements for the United Nations.

In Mali, meanwhile, an adviser with ties to intelligence services allegedly used Claude to develop a platform for monitoring telecommunications. According to the report, it was designed to target 25 million SIM cards and circumvent the requirement to obtain judicial authorization.

Anthropic says it uncovered and shut down all of these different operations between December 2025 and August 2026, primarily by deleting the accounts involved and strengthening its detection systems.

This article was originally published in German.