Explicit std::unique_lock constructor from mutex

I see that the constructor overload that takes a mutex is marked explicit. I don't see the reason to specify it so. I think there is no harm to allow implicit conversion from mutex to a corresponding std::unique_lock. This can actually be convenient. For example, given void func_a(std::unique_lock<std::mutex> lock) and std::mutex mut, we could directly invoke func_a on mut like func_a(mut). What may be the potential harm with respect to the call func_a(mut)? I don't think there is any effect that may surprise the user. So, the question is: what is the rationale behind marking the constructor explicit?