Cognitive Warfare and the Brutalization of the Information Arena

“The real target is the notion of trust.” This is how Marc-Antoine Brillant, head of VIGINUM, sums up what information operations are seeking. Beyond an election or an opinion, what is being targeted is the bond between citizens and their institutions and between citizens themselves: “trust in their representatives, in official statements, in scientific truth, and even citizens’ trust in one another.” Behind disinformation lies a geopolitical intent: to fragment, polarize and weaken social cohesion where it has vulnerabilities. “We are a democracy. Public debate is free and open, therefore it can be exploited, therefore it is instrumentalized.”

The Brutalization of the Information Environment

“The brutalization of the information environment.” Guillaume Kuster, Deputy Director at the Directorate of Communication of the Ministry for Europe and Foreign Affairs, uses these words to summarize what has changed before OSINT Day: not online insults, which he considers almost anecdotal, but the volume of unsubstantiated arguments circulating all the way into international institutions and gaining credibility there. “In international relations, we have moved from the notion of soft power to a notion of openly asserted influence.”

An operation does not necessarily need to convince; it only needs to fuel the divide where it already exists. At the PhilosoFIC roundtable, Émilie Bonnefoy, founder of Open Sézam and a reserve officer at the armed forces headquarters, draws on VIGINUM reports: the red hands at the Shoah Memorial, pig heads placed outside mosques, bedbugs… Information operations always emerge in contexts of social fragility. Adversaries identify the issues that cause fragmentation, primarily racism and insecurity, and make them evolve simultaneously in the physical sphere and in the information sphere, where pre-existing fault lines allow the narrative to spread. “This sociological terrain is highly conducive to cognitive action,” confirms Guy-Philippe Goldstein, a lecturer at the School of Economic Warfare, at the same roundtable.

The logic extends beyond the social sphere. Émilie Bonnefoy returns to the crash of an Indian Rafale at the Indo-Pakistani border, which fueled an information war that, according to her, was taken up by Azerbaijan for two simultaneous objectives: casting doubt on French nuclear deterrence, which the aircraft embodies, and weakening Dassault Aviation at a time when the company was selling Rafales to India. Military, economic and diplomatic spheres. “A holistic approach,” she says, in which “the boundary between information warfare and economic warfare does not exist.”

Kuster, for his part, distinguishes between operating methods according to their origin. “When it comes from Russia, it is generally intended to make people oppose each other quite violently in the political arena, to push the extremes upward. When it comes from the United States, it is more about attacking what makes the European Union an institution.” Here, both public opinion and the instrument itself are targeted: getting people to accept that regulating platforms from Europe amounts to censorship means disarming the only instrument available to the Union to try to clean up this space. Narratives are constructed, he continues, so that European citizens reject regulation themselves. “This threat is not one that can be dealt with hastily,” insists Brillant. “It is a real, profound threat, with a specific purpose: to destabilize democratic functioning.”

AI, a Structural Multiplier

Artificial intelligence did not invent information attacks. It has changed their scale, precision and point of entry. The point of entry first, as far upstream as possible: Tariq Krim, founder of Cybernetica.fr, describes at PhilosoFIC what he calls epistemic warfare, the poisoning of language models before an opinion has even been formed. “The Russians have produced so many fakes about Ukraine that the LLMs that indexed the web have incorporated much more measured views of the danger posed by Russia’s actions.” It is no longer only public debate that is being targeted, but the infrastructure of knowledge that feeds it. “There is a whole process of de-poisoning that needs to be carried out,” he warns.

Precision comes next. In 2024, during the Olympic Games and the snap legislative elections, VIGINUM observed “the first case in France where we saw the use of generative AI by foreign actors to interfere in public debate.” Not a crude deepfake impersonating a candidate, contrary to the expected scenario, but “rather clever, fairly sophisticated things, the creation of visuals with a humorous tone, playing on irony, to accompany very powerful narratives.” The tool is therefore not primarily used to manufacture spectacular falsehoods, but above all to fine-tune the targeting. “Generative AI used intelligently in broad campaigns to carry out very precise targeting of communities,” specifies Brillant, specifically targeting those with a strong capacity for engagement.

Finally, scale, and with it a shift in motive. The AI-generated video announcing a military coup in France, posted online on December 10, 2025 and viewed more than twelve million times, met all the criteria for foreign digital interference: a foreign actor, inauthentic means, manifestly false content affecting fundamental interests. All except the intent. “There was no desire to destabilize France; there was a desire to make money.” During the March municipal elections, the agency saw, for the first time, private operators based abroad setting up operations of the same kind to monetize traffic. The public report on the protection of elections, published on June 11, names the phenomenon: “the rise of an interference economy,” with actors marketing “intermediation services” to carry out interference on behalf of third parties. The volume of these profit-driven operations already presents VIGINUM with a sorting problem: distinguishing, within the mass, what is driven by the lure of profit from what intentionally threatens public debate.

The distribution environment is working in the same direction, without any adversary needing to intervene. “A social network is not a public information service; it is a service with a profit-making purpose,” Brillant points out, seeing this as a systemic pressure distinct from any hostile action. According to the second edition of Arcom’s barometer, published in January 2026, four out of ten French people get their information daily via social networks and 20% use a conversational AI tool for information at least once a week. Marc-Antoine Brillant points out that these tools cannot “really” guarantee information that is “reliable and verified in what they produce.” Models fed with false information thus become the information gateway for part of the public.

Stepping Into the Arena

“Strategic silence, simply saying nothing on the grounds that the debate does not deserve a response, that arrogant posture no longer has a place today,” Guillaume Kuster states. “We have to step into the arena, with the weapons of the mind rather than through brutality.”

The diplomat presents French Response as embodying this move away from passivity. The official X account of the French Ministry for Europe and Foreign Affairs, with a tone that is sometimes fact-checking, sometimes humorous, was launched at the time of France’s recognition of the State of Palestine. “We took a lot of flak from everywhere, from the Israelis, from the Americans. Minister Jean-Noël Barrot felt that there needed to be a way to respond at the same level.” The first message responded to Marco Rubio on the Palestinian issue, “a kind of fact-checking,” Kuster explains. The new diplomatic grammar is openly embraced for what it is: “The world on X is that of a sixteen-year-old American teenager. In terms of tone, that is what we are trying to do, but in a dignified way.” The response to Elon Musk, who accused the British government of fascism, alone generated more than 8.5 million views and around 800 articles in the French and international press.

Kuster cites some 200,000 followers and more than fifteen million monthly views, while France Diplomatie, the institutional account with more than one million followers, generates around five million.

“We need a resolutely offensive response doctrine to impose costs on our adversaries,” says Brillant. Yet no cost can be imposed on an actor whom we refuse to identify: imposing sanctions requires naming them. Kuster dates the turning point to “a small cultural revolution”: the hacking of TV5 Monde, which was clearly attributed, whereas France had previously limited itself to characterization. Regarding information manipulation, the first individuals and entities to be placed under EU sanctions date back to July 2023, with seven individuals and five entities from the RRN/Doppelgänger network. Since October 8, 2024, a dedicated framework has targeted those responsible for destabilizing Russian activities, including coordinated information manipulation. “It works,” says the head of VIGINUM. Six waves of listings later, the regime covers 69 individuals and 19 entities as of April 21, 2026, according to the Council of the EU. Its scope remains limited: what the Union designates are operators, an operating method, an outfit, a relay, not states.

What remains is the gap between what can be measured and what is being targeted; trust has no unit of measurement. Goldstein drew the following operational conclusion at PhilosoFIC: “Cognitive defense will necessarily have a sociological component.” Marc-Antoine Brillant, for his part, refuses to make this solely a matter for the state. If trust is indeed the target, “we all, to some extent, carry within us the answer to this challenge.” His advice: when in doubt, step away from your feed, your conversational agent, and go and verify.