Researchers are finding new ways to use artificial intelligence in forensic investigations, from identifying insect species at crime scenes to organizing complex evidence in laboratories. These advances could make it faster and more accurate to determine the time of death or analyze data, but experts stress that technology should never replace careful human review.
At Louisiana State University, organic chemist Rabi Musah and her team developed a fast method to identify blowfly species from their puparial casings, the small, hardened shells left behind after maggots mature into adult flies. Because blowflies are often the first insects to colonize a body and different species develop at different rates, knowing which species is present helps investigators estimate how long a person has been dead.
Identifying species from casings has long been difficult, especially when DNA has degraded. Musah’s group used field desorption–mass spectrometry to map each casing’s unique chemical fingerprint, then applied a machine learning model to identify species in about 90 seconds, the team reported October 1 in Forensic Chemistry.
The model was trained on hundreds of samples from lab-raised blowflies before being tested on 19 unknown casings from around the country, each identified correctly. Musah believes the method could also reveal if a body was moved or if toxins were present. “These molecules are like a language,” she said. “If you’re listening, there’s all this information you can extract.”
The technique might also aid cold cases. Because some chemicals persist for years, mapping how they change could allow scientists to estimate how long remains have been exposed. “Casings will remain with the corpse,” said Falko Drijfhout, an analytical chemist at Keele University in England. “If investigators find casings from species that live far away, that’s a sign the body was relocated.”
AI’s expanding role in forensic science extends far beyond insect analysis. At the Harnessing AI for Forensic Science Symposium, hosted by RTI International with support from the National Institute of Standards and Technology (NIST) and Johns Hopkins University, researchers and law enforcement officials discussed how AI might help crime labs work more efficiently.
Predictive modeling could analyze past case data to estimate how long new cases will take and help labs allocate staff and resources. Machine learning could also prioritize evidence based on complexity or the likelihood that a sample will yield results.
But mistakes could have serious consequences. If evidence is wrongly classified as low priority, it might never be tested, potentially affecting justice for both victims and defendants. “Any AI system would need proven reliability and robustness before it is deployed,” said Daniel Katz, director of the Maryland State Police Forensic Sciences Division.
AI could also merge results from different evidence types, such as DNA, fingerprints, and trace materials, to identify patterns and suggest leads. Niki Osborne, of The Forensic AI, called this approach a way to “reduce guesswork and build a more responsive, data-driven case management system.”
Still, AI in forensics must meet high standards for transparency and accountability. NIST defines trustworthy AI as valid and explainable. Michael Majurski, a NIST research computer scientist, compared AI to a witness with no memory: “What it says now has no bearing on what it said in the past, and there’s no way to trust its track record.”
To maintain trust, experts recommend detailed audit trails showing how AI reaches conclusions and training analysts to recognize when not to rely on automated results.
As forensic science continues to evolve, both researchers and investigators face the challenge of balancing innovation with reliability. Whether identifying fly casings or analyzing digital evidence, the goal remains the same: to make forensic investigation more accurate and dependable.
