Custom Questions for Each Student Based on Their Submitted Work?

Generative AI has the ability to create questions for students from the syllabus. This is very easy to do: you point the AI at the course documentation and tell it how many and what sort of questions you want. If multiple choice it also provides the answers. Could this also be done for each student, with custom questions based on their individual project work? 

Much has been written about oral examinations as the solution to students cheating. This is traditionally done for project based students, such as those undertaking a PhD. I had to go through a similar process at the end of my MEd, asked questions by a panel, based on what I submitted. However, this would be infeasible expensive for a program with hundreds, or thousands, of students. One option would be to have students question each other (this as been used in the ANU Techlaucher program). Also group projects reduce the workload, while introducing their own problems. 

Recently when attending training for Instructure's Canvas Learning Management System, being introduced at the Australian National University, I asked if each student could be given a separate file during an online test. The idea would be to provide them with a copy of their assignment, then ask them questions specifically about it. At the time I assumed the same set questions would be asked of every student. But it occurs to me now, why not ask them each a custom set of questions based on their work, and not provide a copy of the work to them?