Artificial intelligence is no longer simply a tool for making human work faster. It is becoming an instrument of power, capable of generating propaganda, conducting cyber operations, building surveillance systems and shaping public opinion at a scale that was previously difficult to achieve. Work that once required teams of writers, translators, researchers and media operators can increasingly be accelerated, coordinated and automated. The result is a new kind of geopolitical competition in which influence itself can be industrialised.

A recent threat intelligence report by Anthropic, the AI company behind the Claude models, offers a glimpse of what this looks like in practice. The company disclosed that it had disrupted an operation linked to Politology, the Africa Corps/Wagner influence branch that Anthropic assesses came under the control of Russia’s Foreign Intelligence Service in late 2023. A Russian-speaking operator based in Bangui used Claude to produce daily content for Radio Lengo Songo, a Central African Republic radio station, generating pro-government, pro-Russia and anti-France narratives that were then amplified through Russian state outlets including RT, Sputnik Afrique and TASS, as well as the Russian House in Bangui. The operator reportedly instructed the model to make the material “look like ordinary journalism” rather than AI-generated content, and separately used it to draft forged communications attributed to the Central African Gendarmerie and Ministry of Defence, and to track opposition figures. Notably, Anthropic says Claude refused a request to name real individuals as militants, a detail that speaks to the limits built into the system rather than its capture by it.

“The objective is no longer only to tell people what to believe; it is to obscure who is asking them to believe it.”

The significance of this case goes beyond the volume of material AI can produce. It lies in the possibility of making foreign influence appear local. Traditional propaganda depended on people to write, translate, adapt and distribute messages, a slow and labour-intensive process. AI increasingly performs many of these functions at once. Anthropic describes the technology as a “force multiplier”: operations that once required specialised teams can now be run faster, at greater scale and at lower cost. This changes the economics of influence, but the more consequential shift is in its visibility. A narrative created or coordinated by a foreign actor can now pass through a local radio station, circulate on social media, and reach audiences as though it were an independent domestic view.

Russia has long understood that geopolitical power extends beyond military reach into the information space. The Central African Republic case does not establish that Russia’s entire African messaging strategy is now automated. What it demonstrates is narrower but still significant: AI can be folded into an existing network of media outlets, personnel and political relationships to make foreign messaging faster to produce, easier to localise and harder to trace to its source.

FROM PERSUASION TO SURVEILLANCE
Anthropic’s report also documents a second, quite different case, one that shows AI’s reach extending from persuasion into surveillance. The company says it identified an actor, likely an independent consultant based in Bamako, working with Mali’s state intelligence service, the Agence Nationale de la Sécurité d’État, who used Claude as the primary engineering workforce behind a national surveillance platform named Lakana 360. The system was designed to monitor roughly 25 million SIM cards across Mali’s three mobile network operators, collecting call records, text messages and voice traffic, matching identities across cards, and generating dossiers on individual phone numbers. According to Anthropic, the consultant removed a warrant check from the dossier function, at an operator’s request, circumventing the legal safeguard that would otherwise have required judicial authorisation before such records were disclosed.

The distinction that matters here is what Claude actually did. It was not used to analyse finished surveillance output; it was used to help design and build the software that made the surveillance possible. Anthropic’s ban on the account halted further design assistance but did not disable the system already deployed, which continued to run on other, locally hosted models. The case does not show that the Malian government itself directed the misuse. It shows something arguably more important: that a single individual, working with a state security service, could use a commercial AI model as an engineering workforce to help build surveillance infrastructure at national scale.

Together, these cases point to a challenge broader than disinformation. The same systems that can generate a convincing local narrative can also help build the tools to monitor the populations exposed to it. For African states, the question is no longer only whether a piece of content is true or false, but who produced it, who financed it, who is amplifying it, and whether the apparent local voice behind it is local at all.

WHAT IT MEANS FOR GHANA AND WEST AFRICA

Ghana’s own experience during its 2024 elections offers an early domestic parallel, if not a directly connected one. The Media Foundation for West Africa documented fabricated audio recordings of the two main presidential candidates, while Freedom House reported AI-generated deepfake audio and video alongside coordinated bot activity spreading misleading narratives. What Ghana encountered as isolated synthetic-media incidents could, as the tools mature, become part of far more systematic operations of the kind Anthropic has now documented elsewhere on the continent.

Information does not respect West Africa’s borders, and a narrative built in one country can be adapted and amplified in another within minutes. ECOWAS has historically organised its security cooperation around terrorism, organised crime and border control. The evidence now emerging suggests that information integrity and digital authentication need to sit alongside those priorities, with member states building the capacity to identify coordinated foreign influence, verify digital content and share intelligence on emerging technological threats.

The lesson for West Africa is not that another foreign power has acquired a new propaganda tool. It is that influence, cyber capability and surveillance are converging inside the same technological ecosystem. The states that prepare for that reality will be defending something larger than their information space. They will be defending the institutions and, ultimately, the sovereignty that rests on citizens’ ability to know who is speaking to them, and why.

Joseph McCarthy is an analyst and researcher specialising in governance, security and political transitions in the Sahel. He writes on geopolitics, development and African diplomacy. [email protected]