As climate change threatens agricultural yields, crop diversification, including the reintroduction of underutilized species, has emerged as a proposed solution for resilience. But what determines the extent of environmental change that underutilized crop species can cope with upon their reintroduction? One critical factor is maladaptive plasticity, a theoretical concept suggesting that populations may respond negatively to rapid environmental shifts, at least in the short term. To test this empirically, we conducted a repeated-measures study using historically underutilized Amaranthus species as a model. Upon reintroduction, we observed higher plasticity in vegetative and reproductive traits than in life-history traits. This suggests that while these species adapt to immediate environmental changes through adjustments in growth and reproduction, limited plasticity in life-history traits may constrain long-term adaptation. Our study aimed to (1) identify the environmental factors driving these plastic responses, (2) determine whether these responses are maladaptive, and (3) assess if maladaptive plasticity also occurs within native environments. Using machine learning models, we found that temperature was a primary driver of plastic responses. Generalized Additive Model (GAM) reaction norm analysis further revealed these temperature-induced responses to be maladaptive, suggesting a mismatch between Amaranthus' traits and the new environment. Applying transfer learning to predict responses in native settings, we found similar maladaptive responses to temperature, indicating that maladaptive plasticity may not be limited to non-native environments, thereby complicating climate adaptation efforts. This study underscore the need for detailed trait plasticity assessments before crop reintroduction for agricultural resilience.