The draft bill on forensic DNA phenotyping (FDP) in Germany should take into account the technology's scientific and operational limitations, as well as the social contingencies regarding its use (see Nature 543, 589–590; 2017).

The high probabilities for predicting visible traits and biogeographical ancestry, as emphasized in the bill and in public debate, do not take into account the prevalence of traits in the German population. Prediction probabilities for rare characteristics can drop to less than 50% when adjusted for prevalence. Yet in practice, FDP-guided investigations will focus on rare characteristics: a positive test for dark skin, a rare trait in Germany, could much more efficiently (even if wrongly) narrow down an investigation than could a positive result for the much more common trait of light skin. Minority groups could therefore become over-represented in police investigations.

Biogeographical testing is less reliable for individuals with mixed ancestry or for those from regions that are undersampled in the reference databases. The use of test results may be premised on assumed visible features (C. Phillips Forensic Sci. Int. Genet. 18, 49–65; 2015). Biogeographical ancestry, ethnicity and appearance are therefore at risk of becoming conflated in policing practice.

Germany's police, prosecutors and forensic analysts need more statistical training to assess test results properly. The use of these technologies also requires a balanced framework of governance, including judicial, ethical and regulatory oversight by independent governmental bodies (see go.nature.com/2oq090e).