The U.S. mining industry remains one of the most hazardous occupational sectors, with persistent safety challenges despite regulatory and technological advancements. Traditional safety approaches such as manual inspections, compliance checks and reactive hazard responses often fall short in protecting workers. This study examines how artificial intelligence (AI)-driven predictive risk assessment can strengthen human resource (HR) strategies for occupational safety and workplace accident prevention in the U.S. mining sector. Two linear regression models were developed to support HR-led safety decision-making: an environmental hazard model based on atmospheric variables (methane, CO, temperature, humidity, dust) and a worker accident risk model based on human-centered variables (experience, shift duration, consecutive workdays, training scores). Together, these models formed a unified Total Mine Risk framework validated on U.S. mining datasets. The models achieved 70–76% accuracy, which enables risk detection up to 48 hours in advance, reducing machinery-related fatalities by 24% and projecting annual savings of $300,000. Through providing interpretable outputs, the models allow HR and safety leaders to integrate predictive insights into training programs, fatigue management, workforce scheduling and compliance reporting under Mine Safety and Health Administration (MSHA) regulations. Beyond operational efficiency, the study highlights how transparent, HR-driven analytics can enhance employee trust, build a proactive safety culture, and position HR as a strategic partner in accident prevention. Overall, the findings demonstrate that AI-powered predictive risk assessment is technically and economically viable and an HR-centered innovation for safeguarding employees in high-risk industries.