This commentary critically reviews a recently published discussion between Hahn et al. (2023) and Berger et al. (2023) regarding recommendations for the use of probabilistic genotyping systems in criminal proceedings, in particular the proper understanding for the evidentiary use – across legal systems – of results produced by such systems and the communication of system outputs to the judiciary. We find that the exchange between Hahn et al. (2023) and Berger et al. (2023) reveals a profound divide between diametrically opposed positions, which is symptomatic of a lack of clarity in some quarters of forensic science about the role of expert witnesses and, in particular, novel forms of so-called machine-generated evidence in the legal process. We argue that in order to prevent scientific practices from inappropriately invading judicial territory, lawyers should take a more active role in scrutinising recommendations and position statements published by members of the forensic science community.