This paper presents a comprehensive framework for personalized generative memory models tailored for human-AI co-creation in design tasks. By integrating memory-augmented generative artificial intelligence (GenAI) systems, the proposed approach adapts outputs based on long-term user interaction histories, ensuring contextual relevance and user-specific design alignment. A trust metric-based federated learning (FL) frame- work is introduced to maintain integrity and accountability across distributed data silos while prioritizing user privacy. Additionally, a novel methodology quantifies and optimizes the trade-off between explainability and performance, addressing the need for transparent and efficient AI systems in collaborative de- sign environments. Extensive experiments on synthetic and real- world design datasets demonstrate significant improvements in personalization, trust, and interpretability compared to baseline models. The proposed framework achieves a balanced trade- off, enhancing user satisfaction and system reliability in creative domains. This work provides a scalable, privacy-preserving, and interpretable solution for advancing human-AI collaboration in design tasks.