SMARCB1 is a core unit of the BAF chromatin remodelling complex and its functional impairment interferes with the self-renewal and pluripotency of stem cells, lineage commitment, cellular identity and differentiation. SMARCB1 is also an important tumour suppressor gene and somatic SMARCB1 pathogenic variants (PVs) have been detected in ~ 5% of all human cancers. Additionally, germline SMARCB1 PVs have been identified in patients with conditions as clinically diverse as Rhabdoid Tumour Predisposition Syndrome type 1 (RTPS1), schwannomatosis and neurodevelopmental disorders such as Coffin-Siris syndrome (CSS). RTPS1 is characterized by the occurrence of highly malignant atypical teratoid rhabdoid tumours (AT/RT) affecting mostly infants, whereas SMARCB1-related schwannomatosis is generally diagnosed after the age of 30 and is characterized by benign schwannomas. Patients with germline SMARCB1 PVs and neurodevelopmental disorders do not usually develop SMARCB1-deficient tumours but instead exhibit severe intellectual disability and congenital malformations. It is intriguing how germline SMARCB1 PVs can be responsible for these very different pathologies. However, a network of different factors has emerged that play important roles in this context. Thus, the tumour phenotype associated with germline SMARCB1 PVs is determined by the nature and location of the SMARCB1 mutation and the timing of SMARCB1 inactivation in specific progenitor cells. Biallelic complete loss of SMARCB1 function during a narrow time window of early embryonic development in neural crest cells is essential for AT/RT development. By contrast, hypomorphic SMARCB1 PVs during later developmental stages affecting more differentiated Schwann cell precursors give rise to schwannomas. However, the loss of the wild-type SMARCB1 allele is insufficient for schwannoma growth which appears to be dependent upon concomitant somatic NF2 PVs in patients with SMARCB1-related schwannomatosis according to the four-hit/three-step model of tumorigenesis. In patients with neurodevelopmental disorders such as CSS, germline PVs would appear to cluster within the C-terminal SMARCB1 domain, interfering with the nucleosomal interactions of SMARCB1 but not with its tumour suppressor activity.