Polarized debates, deepfakes, artificially pushed themes on social networks… This Tuesday, January 7, the teams of Viginum, this agency for the fight against foreign interference and the manipulation of information, received a visit from Clara Chappaz, the Minister Delegate for Artificial Intelligence (AI) and Digital Affairs.
“For security reasons,” no journalists were able to attend the trip. The service, attached to the General Secretariat for Defense and National Security (SGDSN), is tasked with detecting and analyzing very specific information manipulation campaigns on online platforms: “those that involve foreign actors and that aim to destabilize the fundamental interests of the nation.”
The day before, a briefing was organized in advance for the press. The opportunity to learn that the minister intends to ensure, during her visit, "the level and robustness of Viginium's resources", in particular by talking with the data scientists and the service's "data lab". For the teams, it was also a question of highlighting the defensive missions of their service, which is often unknown to the general public.
Neither an intelligence service nor a judicial police service
Viginum, described as the "armed wing of the public authorities", or the "French shield against foreign interference", is neither "an intelligence service" nor "a judicial police service that can make requisitions", recalled Marc Antoine Brillant, head of the agency. The entity seeks to be "aware of what foreign adversaries are doing with, each time, the idea of "putting a countermeasure", underlines the head of the department. Its role is therefore not to pronounce on facts, but to identify "the operating methods, the dissemination techniques (of false information on platforms) used by foreign actors, whether state or not", continues the director of Viginum.
The online investigative work is carried out for the benefit of Arcom, and on behalf of the European Commission. Since the DSA, the European regulation on digital services, platforms such as X, YouTube, Instagram and TikTok must in fact respect "reinforced obligations of moderation and deletion of manipulated information". They must also "limit the virality of false information campaigns, undermining our democracies", recalls the minister's entourage. And Viginum's role is precisely to "document and objectify the inadequacies or limitations of these platforms", which will provide material for the ongoing investigations conducted by the European Commission.
Taking stock of the platforms' cooperation
The service has set up "communication channels" with the main platforms such as YouTube, X, Instagram. "We share detection elements with them, and they give us answers based on what they are willing to tell us", explains Marc Antoine Brillant.
One of the challenges of the visit was also to take stock of the platforms' current cooperation, and how they can improve this level of cooperation. "There are questions that arise in terms of access to data, APIs, in particular, and without naming them, there is still a form of heterogeneity (between the different platforms)", explains the minister's entourage. "We expect platforms to cooperate fully with the competent authorities. It is their responsibility under the DSA and the protection of our democracies", declared Minister Clara Chappaz in a press release published on Tuesday.
Over the past 16 months, eight campaigns identified
In concrete terms, Viginum detects "between 230 and 300 phenomena per year", which result in the observation of "several dozen foreign digital interferences". Which ones? "Over the past 16 months, we have denounced in various reports, with our comrades from the Ministry of Europe and Foreign Affairs, eight campaigns that have targeted the French public debate. The two actors (…) the most aggressive (…) (are) unsurprisingly pro-Russian actors and pro-Azerbaijani actors".
The director of Viginum also returned to the three major malicious uses of AI observed recently. Open source generative AI tools are used to generate "credible, very realistic fake content", known as deepfakes. AI also makes it possible to "create inauthentic but very realistic accounts", or even "to massively distribute content on several platforms", he emphasizes.
A certain desire "of our adversaries to dilute manipulated information in opinion"
To detect them, Viginum has equipped itself, since its creation in July 2021, with a "center of excellence in data science and therefore in artificial intelligence». The service has developed tools based on computer programs and statistical calculations. The latter "allow in a debate around, for example, a keyword, a hashtag, to be able to detect the number of bots or trolls". Other in-house tools are used to "go back to the first content distributors." Systems "also make it possible to detect the duplication of content that may have been based on artificial intelligence," explains Marc Antoine Brillant.
But today, the service must face a major challenge: that of "distinguishing what is part of the synthetic debate (created by AI for the purposes of manipulation, Editor's note) from the authentic or natural debate," adds the manager. The head of the service notes a certain desire "on the part of our adversaries to dilute manipulated information in public opinion. And one of the ways to dilute this manipulated information in public opinion (...) is probably, in the long term, to make greater use of AI, i.e. synthetic content."
0 Comments