In their paper they discuss the shortcomings of the current regulatory frameworks in the EU and the US on the use of AI in developing and deploying clinical genomic tools. They took issue with: inadequate coverage of relevant medical devices regulations, use of systematically biased data for device testing and approval purposes, and lack of robust approaches to measure diversity or representativeness of testing and target populations. 

While the upcoming applicable AI laws can address some of the mentioned concerns, still some areas will stand out. Firstly, how to craft an effective disclosure of the limitations of the tests for diverse populations. Secondly, the inconsistency of the data demands with restrictive data protection laws, and lastly, the unintended impact of the liability rules on the deployment of the tools in clinical settings.