AI deepfakes and fake fighters reshape Mali’s information war

The study, Algorithms at War, by the Dakar-based Timbuktu Institute examines how generative AI is being used across Mali’s conflict – including by JNIM jihadists linked to al-Qaeda, their allies in the FLA, the Front de libération de l’Azawad, a Tuareg independence movement, Mali’s transitional authorities and Russia’s Africa Corps.

“Generative artificial intelligence is becoming a new front in this information war in Mali,” Bakary Sambe, director of the Timbuktu Institute and a researcher and lecturer at Senegal’s Gaston Berger University in Saint-Louis, told RFI.

“Everyone is adopting the tool in their own way, depending on their resources, their level of technical sophistication and their objectives. With artificial intelligence on the Malian front, we are moving from fairly rudimentary use by jihadist groups, for example, to a genuine professional machine on the Russian side.”

Malian PM’s visit to Algeria underlines need for mutual security goals

Personalised recruitment

JNIM sits at the rudimentary end of that spectrum.

The Timbuktu Institute identified AI-generated propaganda posters distributed through its media channels, including one promoting real videos filmed during attacks on 25 April that killed Mali’s defence minister Sadio Camara and drove Malian forces and their Russian Africa Corps partners out of the town of Kidal in northern Mali.

AI does not create new divisions in Mali but amplifies those that already exist, Sambe said.

JNIM remains in a phase of limited experimentation, although researchers believe AI could help the group translate propaganda and greatly increase the amount of material it produces.

“The real danger is large-scale personalisation of the profiles targeted for jihadist recruitment,” Sambe added.

Automatic translation could allow a group such as JNIM to produce recruitment messages in around 15 Sahel languages, including Fulani, Bambara, Wolof and Tamasheq, with different messages aimed at different audiences.

“A message about pastoral pride for young Fulani people, a message about social dignity for marginalised urban youth”, Sambe said.

“And with voice dubbing, you can even reach people who cannot read or write, who until now were beyond the reach of written propaganda.”

Insurgents attack multiple towns in northern and central Mali, army says

Fake fighters

FLA independence supporters are using AI differently. Unofficial social media accounts run by members or supporters have created entirely artificial characters to spread their messages.

One example identified by researchers is “Azzghim”, a supposed Tuareg fighter whose fierce criticism of Russia’s Africa Corps has been widely shared online. He is in fact an extremely realistic AI creation, although imperfections in the image reveal that he is artificial.

Such characters can “humanise” a movement, tailor messages to particular audiences, make them easier to spread and help recruitment, the study’s authors said. They also reduce the risks a real fighter would face by appearing publicly.

Russia’s Africa Corps and Mali’s transitional authorities, meanwhile, show a “state-level, professionalised and systematic adoption of generative AI”, the Timbuktu Institute said.

Africa Corps, which succeeded the Russian mercenary group Wagner, has incorporated AI into an already extensive disinformation operation.

Its methods include fake videos, anti-Western deepfakes and generated social media comments, with large-scale operations documented by researchers and European state agencies.

Global jihadism’s growing grip on Africa

Propaganda at industrial scale

“The Africa Corps case is undoubtedly the best-documented case in the entire study,” Sambe said.

Researchers have counted 19 Russian disinformation campaigns across Mali, Burkina Faso and Niger since 2018, he said.

“With AI, Russian disinformation is moving to a completely different scale. We are going from cottage industry to industrial production, with plausible deniability thrown in,” Sambe said.

Some content identified by the Timbuktu Institute has attracted nearly five million views.

It depicts victories by Malian forces over JNIM and FLA fighters and amplifies messages from Mali’s transitional authorities and the three countries in the Alliance of Sahel States – Mali, Niger and Burkina Faso – against France, regional bloc Ecowas and human rights organisations.

Deepfakes and coordinated campaigns were also used after the fighting in Kidal on 25 April to present Africa Corps in a favourable light and avoid the appearance of defeat, Sambe said.

Russian paramilitaries have also used AI to push claims that France and Ukraine supported and equipped JNIM and FLA attackers during the battle, he said. The claims form part of a narrative portraying foreign powers as plotting against Mali while reinforcing official government messaging, Sambe said.

Such content is used for political control and to manipulate information and opinion, diverting people’s attention and misleading them about the real security situation, the institute said.

Its impact is greater because Alliance of Sahel States governments have tightened control over domestic media and cut off international outlets they consider too troublesome, the report added.

Mali’s war on civilians deepens as alliances shift, HRW says

From propaganda to battlefield

AI could increasingly move beyond information warfare into military operations, including drone strikes, researchers warned.

AI can help paramilitary forces identify targets and increase their capabilities. Its use has also led to mistakes and harm affecting communities, including civilians and other non-combatants, Sambe said.

Advanced military uses remained largely prospective in the institute’s written study, but their potential was described as “a serious threat because of its potential use as a force multiplier” – something that could allow armed groups to achieve more with the resources they already have.

The same effect could transform recruitment by allowing jihadist groups to target people individually with highly tailored messages, Sambe said.

“Tomorrow’s jihadism will no longer recruit en masse. It will recruit person by person and tailor its approach to each individual,” he said.

---

This story was based on reporting and interviews in French by David Baché and Christophe Boisbouvier