Several months ago in Niger, the military junta flooded information spaces with accusations against the French military, alleging airspace violations, support for terrorist attacks, release of jihadists, even preparation of a coup d'état. The cost to the attackers: a few thousand euros at most. Costs and impacts on the French side: a damaged reputation despite all debunking efforts, endangered military and diplomatic personnel on the ground (culminating in the ransacking of the French embassy in Niger), and a compromised withdrawal operation from the territory. This puts us at the heart of a new paradigm: one where winning on the ground no longer suffices if you lose in the information space.
When narrative becomes a strategic weapon
"What fundamentally changes today is that we have examples where, even when you win a battle in the real world, if you lose on the information front, you've completely lost," says Stéphane Grousseau, Deputy Director of the Cyber Defence and Intelligence Agency at Sopra Steria. This observation marks a profound break from traditional military doctrines.
One example illustrates this dynamic with chilling clarity. During a recent aerial confrontation, India probably had technical superiority in the air, but Pakistan, likely with external assistance, was faster and more effective in launching its narrative, emerging victorious from the information battle. "The backlash was immediate: Dassault lost more than 3% of its share price within hours, and both private entities and state actors had to mount an information defence," recalls Grousseau. A battle won in the sky, lost on networks, with tangible economic and diplomatic repercussions.
This new reality doesn't simply represent a change of scale in traditional propaganda. It's a change in nature. Information warfare is no longer an optional complement to military action: it has become an essential component, sometimes even decisive.
AI as an amplifier of asymmetric power
Into this already disrupted context, artificial intelligence arrives as a catalyst. "It's a real game changer," affirms Grousseau. "Compared to the other nine times out of 10 when you hear this expression in computing, generative AI truly is one here."
The impact unfolds across several dimensions. Deepfakes transform our relationship with visual and audio evidence. "The level of image generation, audio, and then real-time video completely changes the game. You can no longer believe what you see," he observes. This erosion of trust in visual media once considered most reliable undermines the very foundations of our collective understanding of reality.
But massification perhaps represents an even more insidious danger. "AI allows you to maintain the same narrative whilst generating fundamentally different wording and tone," explains Grousseau. The result: "Instead of two people telling a story, you have 5,000 different sites telling the same thing in diverse forms. This creates an illusion of massive credibility." The old bot farms required thousands of employees. Today, a few people and algorithms suffice to create artificial consensus on a global scale.
The algorithmic war of superpowers
This technological revolution unfolds within a tense geopolitical context. The United States, China, Russia and Europe are positioning their pieces with radically different strategies. "It's a question of culture, doctrine and legal framework," emphasises Grousseau.
Russia often assumes responsibility for its information attacks. The United States benefits from a permissive legal framework that allows "a much wider field of possibility for information attacks". China massively experiments with AI whilst controlling critical supply chains: it can slow down its adversaries by limiting the export of rare earth elements necessary for manufacturing computing chips. Europe, meanwhile, operates within a more measured legal framework with the AI Act.
These doctrinal differences translate into colossal investments. In January 2025, Trump's announcement of a $500 billion investment in AI constitutes, in itself, an information warfare operation. "Major players like the GAFAM are attempting to force the issue through their strike capability," acknowledges Grousseau. Yet he tempers this: "We'll see the emergence of frugal models and specialised AIs that don't depend on ultra-massive infrastructures."
The permanent advantage of attack
In this reconfigured landscape, one reality remains: the fundamental asymmetry between attack and defence. "There is absolutely no doubt: the advantage lies with the attacker," states Grousseau flatly. This law, well known in cybersecurity, applies with multiplied force in the information domain.
"Attack costs little; debunking costs extremely dearly," he continues. Studies show that fact-checking, despite all its efforts, struggles to convince those who have already accepted false narratives. "Showing photographs of a round Earth to a flat-earther won't change their mind." Worse still: energy devoted to defending against one attack is not available to prevent the next. "It's easier to light a fire than to extinguish it," adds Grousseau.
This asymmetry now structures contemporary conflicts. In Ukraine, in Israel, in the Middle East, AI is used massively for reconnaissance, targeting and coordination of operations. But above all, it amplifies the information dimension of every kinetic action. The Ukrainian Spider Web operation against Moscow (a June 2025 covert drone attack using 117 drones smuggled into Russia to strike five air bases and cause an estimated $7 billion in damage) illustrates this synergy: "Although significant, the operation wasn't massive in destruction, but it generated a resonance chamber and violent impact. It affected Muscovites' morale by making them realise that war could impact their daily lives. The operation would have flopped without this information coupling."
Defence: a necessarily hybrid approach
Faced with these protean threats, what defence can be built? For Sopra Steria, the answer lies in a multi-layered system combining technology and human training.
"We seek to equip our clients to be able to respond to information attacks," explains Grousseau. This involves technological observatories capable of detecting campaigns in real time, specific response capabilities to counter rapidly, and above all "compliance engineers" capable of navigating legal and technical complexity.
But technology alone does not suffice. "Security won't come from an 'information iron dome' but from each person's capacity to exercise critical thinking. We must adopt the Zero Trust approach: trust nothing and no one," insists Grousseau. This cognitive defence constitutes the final rampart, the most difficult to build but also the most durable.