Conclusions

This systematic review provides a comprehensive overview of the current applications of AI in forensic personal identification. The findings indicate that AI-based approaches, particularly deep learning models applied to imaging datasets, have demonstrated high predictive performance across several forensic tasks, including sex estimation, human identification, ancestry estimation, and kinship analysis. The increasing availability of digital imaging data, such as computed tomography and radiographic datasets, has played a key role in facilitating the development of AI-driven analytical methods in forensic research.

Despite these promising results, important challenges remain. The predominance of population-specific datasets, methodological heterogeneity across studies, and the limited availability of externally validated models highlight the need for more standardized research protocols and multi-population datasets. In addition, ethical and legal considerations related to transparency, interpretability, and the potential forensic use of algorithmic outputs must be carefully addressed before widespread implementation in medico-legal practice.

Overall, AI should be considered a complementary tool capable of supporting forensic experts rather than replacing human expertise. Future research should focus on improving model transparency, expanding population diversity in training datasets, and developing standardized validation frameworks to ensure the reliability and forensic applicability of AI-based identification systems.