Students' physical and digital lives are increasingly entangled. It is difficult to separate students' "digital" well-being from their offline well-being given that artificial intelligence increasingly shapes both. Within the context of education's fiduciary and moral duty to ensure safe, appropriate and effective digital learning spaces for students, the continuing merger between artificial intelligence and learning analytics not only opens up many opportunities for more responsive teaching and learning but also raises concerns, specifically for previously disadvantaged and vulnerable students. While digital well-being is a well-established research focus, it is not clear how AI-Powered Educational Decision Support Systems (AI-EDSS) might impact on the inherent, situational and pathogenic vulnerability of students. In this conceptual paper, we map the digital well-being of previously disadvantaged and vulnerable students in four overlapping fields, namely (1) digital well-being research; (2) digital well-being research in education; (3) digital well-being research in learning analytics; and (4) digital well-being in AI-informed educational contexts. With this as the basis, we engage with six domains from the "IEEE standard 7010-2020--IEEE Recommended Practice for Assessing the Impact of Autonomous and Intelligent Systems on Human Well-Being" and provide pointers for safeguarding and enhancing disadvantaged and vulnerable student digital well-being in AI-EDSS.