In the context of growing demand for privacy-preserving data sharing, k-anonymity has emerged as a widely adopted anonymization technique to protect individual privacy while preserving data utility. However, the generalization processes inherent to k-anonymity often lead to substantial information loss and reduced classification accuracy. To address this trade-off, this paper proposes a novel hybrid optimization model combining Artificial Rabbits Optimization (ARO) and Arctic Puffin Optimization (APO). The proposed approach utilizes a multi-objective fitness function that simultaneously minimizes information loss (IL), preserves classification accuracy (CA), and satisfies the k-anonymity constraint by minimizing violations (M). In the hybrid framework, APO is employed for global exploration to escape local optima, while ARO provides efficient local exploitation of promising solutions. The hybrid model integrates both algorithms in parallel and periodically exchanges their best individuals to enhance convergence and diversity. Experimental evaluations on six real-world dataset – including census, healthcare, and demographic data-demonstrate that the proposed model significantly reduces information loss while maintaining high post-anonymization accuracy. The results show its robustness and adaptability across datasets of varying sizes and complexities, highlighting its potential application in secure and intelligent data-sharing systems.