Replicable Monte Carlo simulation with OpenMP

Here is what GFortran documentation says:

Note that in a multi-threaded program (e.g. using OpenMP directives), each thread will have its own random number state.

Does that mean GFortran’s RNG is thread-safe?