In this episode, we dive into the emerging threats facing AI systems in medical diagnostics, particularly in computational pathology. As AI models become integral in supporting cancer diagnostics, adversarial attacks—subtle, malicious modifications to data—pose a significant risk, potentially leading to misdiagnoses. We'll explore why specific AI models, like CNNs, are vulnerable to these attacks while newer models, such as Vision Transformers (ViTs), show greater resilience. Join us to uncover how adversarial robustness could be a game-changer in making AI safer and more reliable for medical use.