We present a forensic analysis of the generation pipeline behind a recent AI driven influence campaign. We introduce PROPAGIA, a corpus of 2,646 propagandist French articles from the Storm-1516/CopyCop campaign disclosed by VIGINUM and INSIKT GROUP in 2025. For comparison, we rely on SIPA, a corpus of human-written French mainstream press from the same period. Using topic modeling, vagueness and sentiment analysis, we first isolate persuasion techniques characteristic of propaganda, with PROPAGIA far exceeding SIPA in vagueness, subjectivity and negativity, and citing fewer sources. We then find prompt instruction leaks on 50 of the 84 PROPAGIA websites, including a verbatim ten-point editorial specification accounting for several of these differences, together with high cross-article redundancy. Finally, we show that rewriting-based detection supports INSIKT GROUP’s attribution to the Llama 3 family, but also suggests the involvement of Mistral-family models.

Relation extraction is a subtask of natural langage processing that has seen many improvements in recent years, with the advent of complex pre-trained architectures. Many of these state-of-the-art approaches are tested against benchmarks with labelled sentences containing tagged entities, and require important pre-training and fine-tuning on task-specific data. However, in a real use-case scenario such as in a newspaper company mostly dedicated to local information, relations are of varied, highly specific type, with virtually no annotated data for such relations, and many entities co-occur in a sentence without being related. We question the use of supervised state-of-the-art models in such a context, where resources such as time, computing power and human annotators are limited. To adapt to these constraints, we experiment with an active-learning based relation extraction pipeline, consisting of a binary LSTM-based lightweight model for detecting the relations that do exist, and a state-of-the-art model for relation classification. We compare several choices for classification models in this scenario, from basic word embedding averaging, to graph neural networks and Bert-based ones, as well as several active learning acquisition strategies, in order to find the most cost-efficient yet accurate approach in our French largest daily newspaper company’s use case.

De nombreuses méthodes d’extraction et de classification de relations ont été proposées et testées sur des données de référence. Cependant, dans des données réelles, le nombre de relations potentielles est énorme et les heuristiques souvent utilisées pour distinguer de vraies relations de co-occurrences fortuites ne détectent pas les signaux faibles pourtant importants. Dans cet article, nous étudions l’apport d’un modèle de détection de relations, identifiant si un couple d’entités dans une phrase exprime ou non une relation, en tant qu’étape préliminaire à la classification des relations. Notre modèle s’appuie sur le plus court chemin de dépendances entre deux entités, modélisé par un LSTM et combiné avec les types des entités. Sur la tâche de détection de relations, nous obtenons de meilleurs résultats qu’un modèle état de l’art pour la classification de relations, avec une robustesse accrue aux relations inédites. Nous montrons aussi qu’une détection binaire en amont d’un modèle de classification améliore significativement ce dernier.