The article by Durán and Jongsma attempts to establish philosophical grounds for trust in medical artificial intelligence (AI) by positing concerns about opacity, ‘black box’ algorithms and epistemic justification. However, their argument conflates distinct technical concepts, mischaracterises both machine learning (ML) and statistical inference and fails to engage with established methodological frameworks used in safety-critical domains. Specifically, the article erroneously treats AI and ML as monolithic, assumes that unexplained correlations equate to epistemological opacity and obscures well-developed approaches in economics, biostatistics and epidemiology for handling uncertainty and causal inference. By comparing the authors’ claims to accepted practice in econometrics and clinical statistics, this response shows that the article’s foundational propositions are technically unsound and ethically unhelpful. A more productive ethical analysis of medical AI should foreground rigorous validation, transparent performance metrics and appropriate frameworks for causal inference and accountability. This response proposes alternative criteria for evaluating trust and ethical deployment of AI in medicine.