Innovación en la identificación de riesgos medioambientales de los SG-SST mediante el uso de inteligencia artificial

Abstract

A bibliographic and documentary review was conducted examining the incorporation of Artificial Intelligence (AI) techniques into Occupational Health and Safety Management Systems (SG-SST) for the early detection of environmental risks. Several approaches based on supervised and unsupervised machine learning, such as convolutional neural networks and clustering algorithms, were identified as effective for detecting patterns of pollution, variations in suspended particles, and thermal anomalies that would otherwise go unnoticed or require extensive manual processes with conventional methods (Tang et al., 2024; Fazli et al., 2025). In a context where the complexity of industrial and extractive environments is growing xponentially, these tools enable continuous analysis of time series and real-time data streams, reducing both error margins and false positives (Abiodun et al., 2023). In the current technological landscape, the integration of Internet of Things (IoT) sensors and unmanned aerial vehicles (drones) provides occupational safety management systems (SG-SST) with multisensory information sources that feed AI cloud platforms. This integration allows the development of interactive dashboards capable of geospatially representing pollutant distribution and issuing automatic alerts when critical thresholds, previously defined according to regulations, are exceeded (AutCon, 2024; Mariano, Almada & Dutra, 2024). Furthermore, these systems not only learn from new field data but also adapt their models through continuous feedback, reinforcing a culture of prevention and enabling proactive planning of mitigation measures (Xu & Saleh, 2020). This literature review addresses the integration of AI into SG-SST for the early detection of environmental risks. Applications of machine learning, computer vision, and time-series modeling combined with IoT sensors and drones for real-time data collection were identified from scientific literature, technical reports, and case studies. The findings show that these tools reduce error margins and false positives, enhance predictive capacity, and support timely, data-driven decisionmaking. At the same time, the review highlights the need for ethical and regulatory frameworks that ensure algorithmic transparency, as well as participatory processes that involve worker validation and training. Thus, AI emerges not merely as a monitoring resource but as a transformative element in organizational culture and environmental management. This research offers a methodological framework that can be replicated across multiple sectors, from construction to mining, and serves as a starting point for the development of future SG-SST standards grounded in evidence and sustainability. As emphasized by Assies (2000) and Grimson (2011) in the field of epistemic justice, acknowledging workers’ experience and local realities in model calibration ensures fairer and more resilient risk management, capable of balancing productivity demands with the protection of the environment and occupational health.