The struggle for influence in Mali no longer relies solely on military statements and recordings of armed commanders. Synthetic personalities, AI-generated images and messages that can be produced in multiple languages are opening another arena in the contest for audiences and control of the narrative.
In a report titled Algorithms at War: How Generative AI Is Reshaping the Malian Conflict, the Timbuktu Institute warned that the country could become a testing ground for AI-assisted information warfare. It distinguishes, however, between observed uses and capabilities that remain hypothetical. The institute published its official overview on September 8, 2026.
The report examines the al-Qaeda-linked Jama’at Nusrat al-Islam wal-Muslimin (JNIM), content supporting the Azawad Liberation Front (FLA), and Russia’s information operations. These actors differ in identity and objectives; their use of similar tools does not place them within a single political or ideological project.
According to the institute’s overview, JNIM’s use remains limited, while pro-Azawad content appears more decentralised. The researchers identify Russia-linked campaigns as displaying the most sophisticated application of the technology.
For JNIM, the analysis begins with visual material that researchers assessed as showing signs of AI generation. In their view, the significance extends beyond individual posters to the possibility of making propaganda cheaper to produce and faster to distribute.
TV5MONDE’s coverage of the report highlights the potential for automated translation to reach wider audiences speaking the languages of the Sahel. Audio and video production could also reach people who do not follow written publications.
These possibilities do not establish that JNIM has already built a comprehensive automated recruitment system, or that it can achieve consistent quality across local languages. This part of the report assesses what the technology might enable as its use expands.
Regarding the FLA, the analysis focuses on accounts featuring AI-generated personalities that promote narratives favourable to Azawad. Such figures can appear familiar and approachable, giving political messages a personal voice and potentially creating an impression of broad popular support.
Attributing this content directly to the FLA, however, is different from identifying messages that support it. The tools are available to sympathisers and individuals who can create and publish material without organisational instructions. Establishing who operates and finances the accounts is therefore essential before assigning responsibility.
The report also looks beyond local actors. Its official overview points to Russia-linked campaigns combining generated content, fake accounts and methods designed to expand their reach, citing African Initiative and the “AI-Freak” operation. These findings remain the institute’s assessment.
The developments raise a challenge that goes beyond identifying fabricated images: how can audiences distinguish genuine support from manufactured endorsement, or a real speaker from a personality that exists only on screen?
A large volume of content does not prove successful recruitment, just as view counts alone do not measure political influence. But lower production costs make it easier to flood platforms with repeated messages, increasing the verification burden on journalists, researchers and the public.
In a country where armed actors compete over territory and legitimacy, understanding who is speaking, on whose behalf and with which tools is becoming essential to understanding the war itself.
