Quantum machine learning (QML) systems face unique challenges from both intrinsic quantum noise and adversarial threats. In the context of QML systems, bit-flip errors can occur due to imperfections in quantum gates, stray electromagnetic fields, or thermal fluctuations. These errors are particularly damaging in noisy intermediate-scale quantum (NISQ) devices, where error correction is either absent or limited. In this paper, we provide a structured overview of error types, theoretical frameworks for quantifying errors and robustness of quantum machine learning models, experimental results, and strategies to model and leverage these noises.