Toward Trustworthy AI in Retail: A Case Study of Vending Machines in Public Environments

Artificial intelligence (AI) systems are becoming increasingly embedded in physical retail environments, enabling real-time personalisation through face and behaviour analysis. This paper critically examines the ethical implications of one such system: an intelligent vending machine (VM) infrastructure deployed in 30 real-world locations across Italy. This infrastructure uses computer vision and machine learning to adapt promotional content based on inferred age, gender and interaction patterns. While such systems are technically effective and privacy-preserving, they raise important concerns around fairness, transparency, consent, and profiling. Drawing on the EU's Ethics Guidelines for Trustworthy AI and the ALTAI framework, we evaluate the VM system against core ethical principles, particularly those of justice and explicability. Key challenges identified include demographic bias in personalisation, the absence of meaningful user awareness and implicit profiling in public spaces. Based on our findings, we propose practical enhancements to improve fairness-aware design, embedded transparency mechanisms and symbolic consent models. Our work provides a field-tested ethical assessment of AI in physical retail and offers actionable strategies for aligning real-world deployments with responsible AI standards.