How would they know for sure if the submitted code was written by a human or AI? If they had a “no-AI” policy, there would be no way to enforce it.

The policy makes sense to me given the security concerns for the project.