5 min read
Incident Reports An NHS trust confirmed adversarial perturbations applied to medical images caused systematic misclassification by its AI diagnostic system, resulting in incorrect preliminary diagnoses.
adversarial examples
5 min read
Research Vision-language models are highly susceptible to adversarial image perturbations, with attacks transferring across models (GPT-4V, Gemini Pro, LLaVA) at 43-74% success rates.