Background - Reported cases of exposure to potentially toxic agents are collected in a national database and average 200,000 cases annually, of which 75,000 to 80,000 are symptomatic. These data are currently used to document signals from local institutional partners (hospitals, regional health agencies), national partners (health agencies), or international partners (Rapid Alert System for Food and Feed [RASFF]). The toxicovigilance network aims to develop an automated detection program for unusual events in order to identify as early as possible signals posing a potential threat to public health. To meet the objective of this program, it is necessary to compile an inventory of methods for the surveillance and automated detection of unusual events that can be used in the context of toxicovigilance. Methods - A literature review was conducted using Scopus® and PubMed®, supplemented by grey literature and by consulting information available on the websites of vigilance systems. Results - The most commonly used methods, particularly for pharmacovigilance, are measures of disproportion. Some of these are used for routine detection activities. Signal generation criteria vary little across systems, but these systems have implemented data filtering strategies—either pre- or post-analysis—to reduce the number of generated signals and improve their prioritization. These signals are then forwarded to a committee of experts for clinical and epidemiological evaluation, and potentially for inclusion in the patient’s medical record. Nevertheless, there is interest in other approaches, such as methods for monitoring time-series data or symbolic methods previously used in pharmacovigilance. These methods allow for the extraction of association rules between one or more drugs and one or more side effects, with the possibility of integrating other variables such as demographic data. The development of specific algorithms based on empirical probabilistic approaches has also been the subject of recent research. Conclusion - The signal detection methods used in pharmacovigilance are of definite interest for the automated signal detection program in toxicovigilance. A preliminary step in the development of these approaches requires studying the statistical quality of the data and identifying the detectable events to be detected based on the available data.