AI tools are already part of how students study. The question for educators isn't whether to engage with that, but how to do it without undermining genuine learning.
Banning AI tools outright is increasingly unrealistic, and treating them as unquestionably safe is naive. The more useful question for bioscience education specifically is: which uses build genuine understanding, and which just outsource the thinking a student actually needs to do themselves?
Used well, AI tools are strong at generating practice questions on a specific topic, explaining a concept a different way when the first explanation didn't land, and helping structure a revision plan around a specification. Used this way, the student is still doing the cognitive work of recall, application and self-testing, just with a more responsive resource than a static textbook.
The risk is highest when a tool is asked to produce the actual output a student needs to learn to produce themselves: a full essay, a lab report discussion section, or a worked solution copied rather than attempted first. The output can look convincing while the underlying skill never actually develops, which surfaces later, in an exam or viva, at a much worse moment than during revision.
Before using an AI tool for a task, it's worth asking: am I using this to check or extend my own attempt, or to avoid attempting it at all? The former builds skill; the latter defers a problem rather than solving it. This is a distinction most students can apply themselves once it's named clearly, rather than requiring constant policing.
Rather than only writing AI-proof assessment (an increasingly difficult arms race), it's often more productive to teach explicitly *how* to use these tools as a study aid, modelling good and poor uses directly, since students are using them regardless of institutional policy, and the quality of that use is heavily shaped by whether anyone actually taught it.