The disclosure appears in Anthropic’s September 2026 threat intelligence report, “Detecting and countering misuse of AI,” published on September 10, 2026, which documents cases of AI misuse the company disrupted between December 2025 and August 2026.
A Network of 70 Fake Newsrooms
According to the report, the network operated 70 fake news websites designed to look like independent local newsrooms, each paired with matching X accounts and amplified by more than 250 inauthentic commenting accounts — many using AI-generated profile photos.
Anthropic traced the operation to LKM Company, a France-based digital advertising agency, and said it used Claude to both write original articles and rewrite real news stories from legitimate journalists into politically slanted versions tailored to different national audiences.
The network published at least 8,913 articles in about 20 languages, targeting audiences across six continents, according to the report.
The DRC: The Network’s Biggest Country-Specific Target
Within that global network, the DRC received outsized attention. Anthropic’s investigators found 318 articles specifically focused on the Democratic Republic of Congo — the largest country-specific cluster of content identified in the entire operation.
The stories, the report says, “typically supported the DRC government’s stance,” with a specific focus on regional mineral deals and tensions with Rwanda.
The DRC-focused page became the network’s most-shared and most-popular property during its early weeks, and the majority of fake follower accounts acquired in that period were tied to DRC audiences, according to Anthropic. An X account presenting itself as a Congolese civilian “digital army” was also observed following several of the network’s accounts.
Anthropic pinpointed September 11, 2025, as a clear signal of coordinated activity: multiple sites in the network published nearly identical articles about the DRC-Rwanda conflict within three minutes of each other, with the tone adjusted for different regional audiences while distribution was coordinated simultaneously across the network’s X accounts.
Anthropic, while linking the network to an “influence-as-a-service” model, said it had banned the accounts associated with the network and developed new detection measures based on the operation’s behavioural signatures to identify similar activity in the future.
Kigali Has Long Rejected the Kinshasa Narrative
The findings land against the backdrop of a conflict in which Rwanda has repeatedly and publicly rejected the characterisation of itself as the aggressor or as a beneficiary of Congolese mineral wealth — the same narrative elements the fake network was found to be amplifying.
In a February 3, 2025, interview with CNN’s Larry Madowo, at the height of the fighting between the Congolese army and AFC/M23 rebels, President Paul Kagame said the war in eastern DRC “was not started by Rwanda; it was started by the DRC.” He argued that the M23 rebels fighting in the region are Congolese, not Rwandans, and are “fighting for their rights” after years of persecution and failed negotiations with Kinshasa.
He linked Rwanda’s tension with DRC to Kinshasa’s continued tolerance of the Democratic Forces for the Liberation of Rwanda (FDLR), a militia group formed by individuals involved in the 1994 Genocide against the Tutsi, which operates freely in the DRC with government support.
Kagame also dismissed accusations, frequently made by Congolese officials and echoed in some international reporting, that Rwanda profits from DRC’s mineral resources.
“I invite you to come and see where we mine coltan,” he told CNN, arguing that Rwanda has its own mineral deposits and does not depend on Congolese resources.
Kagame: ‘Creating Problems and Blaming Rwanda’
A year later, on February 5, 2026, during the opening of the 20th Umushyikirano National Dialogue Council in Kigali, the President maintained that such claims were illogical and deliberately ignored the real drivers of insecurity in the region.
“The war in DRC was not started by Rwanda. It began somewhere else, not in Rwanda,” he said. “The real issue Rwanda faces involves the Interahamwe, who are tied to the ideology of genocide and continue to plot, saying, ‘We are coming to finish what we started.'”
“Creating problems and blaming Rwanda is all that the international community does to Rwanda,” he added. “They treat Tshisekedi [President Felix Tshisekedi] as a spoiled child. There is no single blame that goes to the people who created this problem.”
What It Means
The Anthropic report offers rare, concrete evidence of how generative AI is being deployed at scale in geopolitical influence operations — and shows that African conflicts are not peripheral to these campaigns but central targets. The DRC-Rwanda narrative was the single most heavily targeted country-specific theme in a network that reached six continents.
For Rwanda, the findings appear to corroborate a long-held complaint: that its portrayal in international and regional media as an aggressor and mineral profiteer has been deliberately manufactured and amplified. For Anthropic, the case adds to a growing body of evidence that AI models can be weaponised for state-aligned messaging — and that detecting and disrupting such operations is becoming a core responsibility for the companies that build them.
