Contemporary ethical guidelines for Artificial Intelligence (AI) focus almost exclusively on the deployment and usage of technologies based on machine learning methods. They align with the interests and values of the producers of automated solutions, largely based in the Global North and in some emerging countries. In multi-disciplinary perspective that takes into account inputs from digital social sciences, we propose a socially responsible and environmentally sustainable ‘end-to-end approach’ to AI ethics. Its purpose is to expand the scope of current research to include a systematic study of the social, economic, and environmental contexts of AI production, considering in particular the extraction of natural resources and the data work needed to train, validate, and impersonate AI systems. An end-to-end approach requires promoting the autonomy and advocating the dignity of the lower-level workers mostly located in Global South countries.