AI is recognized as a powerful tool for addressing the complexities of 6G technology. However, trust in artificial intelligence (AI)’s decision-making processes must be established to ensure its successful and responsible integration. A practical trustworthiness evaluation approach is proposed for AI-driven 6G applications, explicitly focusing on automatic modulation recognition (AMR). This approach centers on three core attributes: data robustness, ensuring the reliability of the models despite variations in input data; parameter sensitivity, where minor adjustments to model settings should not lead to drastic performance changes; and security against adversarial examples, demonstrating the resilience of the models to intentionally crafted malicious inputs. This focus on trustworthiness evaluation contributes to developing trustworthy AI solutions, facilitating AI’s secure and dependable deployment in future wireless communication systems.