Welfare states across the world increasingly experiment with the use of big data and algorithms in the name of efficiency gains and fair decision-making. However, recent public scandals in various countries show persistent problems with how digital welfare states operate at the cost of vulnerable populations. Problems include systemic forms of discrimination, increasing levels of surveillance, stigmatization, and restricted access to public benefits. Rather than viewing these problems as instances of technical implementation hurdles, we argue that digital welfare states currently operate in an institutional void in which legal, ethical, and quality procedures are lacking or ill equipped to address new challenges posed by digital technologies. Based on a secondary analysis of documents, we show how this institutional void empirically manifests itself in the Dutch welfare state by zooming in on two subcases that sparked public controversy in recent years: the childcare benefits scandal that focused on fraud detection and the Top 400/Top 600 that was set-up with the aim of crime prevention. By analyzing these cases, we show how sphere transgressions – understood as the encroachment of digital logic into the sphere of social welfare – can have detrimental consequences for citizens when there is an institutional void. We end with reflections on how to fill the current institutional void and identify ‘soft signals’ that could be used as pointers to recognize the potential undesirable consequences of new sphere transgressions.