Digital education is rapidly expanding in primary school contexts, creating both opportunities and risks for vulnerable learners, particularly students with autism spectrum disorder (ASD). While existing research highlights increased exposure to online risks, empirical evidence on how protection is shaped across stakeholders and institutional settings remains limited. Addressing this gap, a mixed-method approach is employed to examine the mechanisms underlying child protection in digital education. In the first stage, semi-structured interviews were conducted with 37 key stakeholders, including teachers, parents, and school administrators, to explore experiences and challenges related to digital safety. Insights from this qualitative phase informed the development of an integrated framework capturing ASD-related digital vulnerability (ASDV), perceived digital risk (PDR), school safeguarding systems (SSG), adult protection capacity (APC), and home–school collaboration (HSC). In the second stage, the proposed model was tested using survey data from 360 respondents (180 teachers and 180 parents) and analyzed through partial least squares structural equation modeling (PLS-SEM). The results indicate that protection effectiveness in digital education (PECP) is jointly shaped by multi-level factors. Student vulnerability influences outcomes indirectly through PDR, while SSG exert both direct and indirect effects by strengthening adult capacity and collaborative practices. Multi-group analysis further reveals that teachers rely more on institutional mechanisms, whereas parents place greater emphasis on relational and collaborative processes. By developing and empirically validating a whole-school protection framework, this study clarifies how institutional safeguards, adult capacity, and HSC jointly shape digital protection for primary school students with ASD. It also reveals stakeholder-specific mechanisms by showing that teachers rely more on institutional safeguards, whereas parents place greater emphasis on collaborative processes.