The long-held belief in the uniqueness of fingerprints, often considered a cornerstone in forensics and biometrics, is being challenged by a study from Columbia University. The research, utilising an artificial intelligence (AI) tool, suggests that fingerprints may not be as distinct as conventionally thought.

Columbia University’s AI tool was trained on 60,000 fingerprints, and its outcomes questioned the traditional understanding of fingerprint individuality. The AI, with 75-90% accuracy, could identify whether prints from different fingers belonged to the same person. Surprisingly, the tool seemed to focus on the orientation of ridges in the center of a finger rather than the minutiae (ridge endings and forks) typically used in forensic analysis.

Despite the promising results, researchers, including Prof Hod Lipson, admitted uncertainty about how the AI achieved its conclusions. The findings could impact both biometrics, where a specific finger is used for device unlocking, and forensic science. The tool might offer leads in forensic investigations by connecting fingerprints from different crime scenes to the same individual.

However, experts urge caution, emphasising the need for further research. Graham Williams, a professor of forensic science, noted that while fingerprints are not definitively proven to be unique, the AI tool’s limitations must be considered. The study is expected to fuel discussions about the potential impact of AI on forensic practices and biometrics.

The Columbia University team acknowledged the need for more extensive research, highlighting that their tool is not currently suitable for courtroom evidence but could assist in generating leads for forensic investigations. The study will be published in the journal Science Advances.