In some of our programs on vhlcentral there is an instructor-facing feature that uses AI to assist instructors to comment on students' open-ended responses to short answer essay questions. The instructor has to opt-in to use the AI functionality. No student data of any kind is transmitted to the AI model and we do not use the AI capability for any kind of grading of the response – it provides feedback that the instructor can choose to use, edit, or delete. before passing the feedback on to the student (human in the loop).
We use an enterprise version of a commercially available LLM for this service. It does not “remember” any input or store the input from students or instructors. No input is used to train the LLM utilized for this service.