Based on the theoretical approaches of social capital and institutional trust, this paper seeks to identify contextual factors and conditions behind teacher behaviours which aim to alter the results of standardised tests in the Italian low-stakes accountability system. Numerous studies report significant factors associated with student cheating, but research into the factors of teacher-led opportunistic actions is scarce. Logistic regression models with fixed-effects at classroom level, with interaction terms, were carried out to identify factors increasing the likelihood of teacher misbehaviour. Models included approximately 79,100 primary, lower and upper secondary classrooms. Indicators of teacher cheating were estimated through algorithms based on suspicious answer strings from standardised tests. The results suggest that teacher cheating may be understood as a form of support for the most vulnerable students, since it is, to a greater extent, found helping low-income students, grade-retained students, as well as students in socially homogenous school settings. The findings also reveal that teacher cheating is consistently related to collectively share non-civic-minded behaviours and practices undertaken by teachers, which do not match legal requirements, such as within-school social segregation and exclusion of students from tests. Heterogeneous effects show that, even in classrooms with external controllers, the lower the civic capital in a school, the more misbehaviour are found. Relevant implications for research, social theory and policy are discussed.