Recent advancements in artificial intelligence, particularly in speech technologies, hold significant potential for improving the health and well-being of older adults by offering noninvasive, accessible, and scalable solutions to increase social engagement, assist with daily activities, and diagnose disease. However, the development of AI models has been almost exclusively based on English-language data sets, marginalizing the over 1 billion older adults worldwide who speak non-English, low resource languages. This lack of linguistic inclusivity restricts their access to innovations, contributes to delayed diagnoses and reduced quality of care, and exacerbates existing health care inequities. Here we highlight the urgent need to curate speech data sets in low resource languages by prioritizing community agency, ensuring equitable distribution of benefits, and establishing sustainable pathways for long-term participation and empowerment, with the aim of advancing inclusive and equitable speech-based health care tools for older adults.