Prostate cancer is among the most common cancers worldwide, with ~1.5 million new diagnoses globally every year. The sheer mass of data becoming available on prostate cancer, as well as other types of cancer, is increasing exponentially. The growth of digital pathology has particularly sparked interest in developing artificial intelligence (AI) approaches to data synthesis to predict cancer grade and outcomes in men with prostate cancer. Progress has been made in this field, particularly in applications for diagnosis, prognosis and inferring molecular alterations, but several challenges remain. Variability in tissue processing and scanning contribute to dataset heterogeneity. The absence of well-annotated, multi-institutional databases hinders AI model development and generalization of model performances across clinical settings. Regulatory frameworks for AI-driven diagnostics remain nascent. Moreover, bias in training datasets skewing against under-represented demographic groups poses a fundamental challenge to developing equitable models. By mapping contemporary evidence around each of these hurdles and identifying tangible interventions, we can advance AI-augmented digital pathology towards reliable and generalizable tools to improve prostate cancer care.