A sophisticated digital network used fake news sites, artificial intelligence, and bot accounts to manipulate discussions on DRC, Rwanda, and critical minerals, according to a report by Anthropic.
Digital disinformation has evolved far beyond isolated posts and viral rumors. Operations now leverage fully integrated infrastructures combining synthetic news portals, sockpuppet accounts, automated text rewriting, and coordinated distribution networks.
In its September 2026 report, AI research company Anthropic revealed it had detected and blocked a network misusing its Claude model to generate and refine targeted content across several countries. The Democratic Republic of Congo (DRC) emerged as a key focal point, particularly regarding ongoing tensions with Rwanda and natural resource governance.
Fabricated Outlets Posing as Independent Media
The campaign relied on approximately 70 websites designed to mimic legitimate news outlets. These portals featured standard editorial categories, professional layouts, and fictitious journalist bylines intended to build public trust.
By disguising promotional or biased messaging as local coverage, the network attempted to pass off coordinated narratives as genuine reporting. Media analysts stress the importance of verifying publisher ownership, author credentials, and underlying sources before treating online coverage as authentic.
Reframing Legitimate News
Rather than relying purely on fabricated stories, the operation frequently hijacked real news reports. Original coverage from mainstream outlets was rewritten, stripped of context, or given a distinct political spin to align with the network’s objectives.
Key themes included regional agreements over mineral resources and diplomatic relations between Kinshasa and Kigali. Anthropic noted that its investigators did not definitively attribute the operation to any specific government entity.
Manufactured Consensus
To amplify its reach, the campaign deployed more than 250 fake user profiles on social media. These accounts routinely shared, commented on, and liked content produced by the network’s pseudo-media outlets, creating the illusion of organic grassroots engagement.
This tactic risks distorting public discourse by making specific viewpoints appear widely held, when in reality their visibility is driven by coordinated automation. Analysts remind readers to cross-check sensitive stories across multiple reputable sources before sharing content online.
(Source: “Countering misuse of AI” report, Anthropic, September 2026)