Interpersonal violence is a relevant concern in care for homeless elderly individuals, LGBTQIAPN+ persons, victims of child abuse, domestic violence, gender violence, or racial violence, among other groups with high social vulnerability. The situation is complex and raises various doubts among healthcare professionals. However, an artificial intelligence (AI) system with a large language model (LLM) and adequate machine training can promote improvement in the health of these populations through humanization of their care. This study thus seeks to implement an AI center and validate an LLM protocol to assist primary care professionals (physicians, nurses, psychologists, social workers, and others) with involvement in surveillance and decision-making on complex cases of violence against these populations, who will be the end beneficiaries. The project is expected to promote prevention, monitoring, analysis of suspicious images for early risk detection, assistance for diagnosis, recommendations of interventions, reduction of exacerbations, compulsory notification, socioemotional support, and referrals according to each case’s specificity. At present, there are no records of LLM in Brazil for this purpose, pointing to the idea as innovative and contributing to the health system in Brazil and in other countries, since the project aims to promote efficiency in data processing, improved understanding of patterns of violence, and generalization to different contexts