Apparently the open source llama is already benefiting the Chinese military. The technology is already out there. I'm sure other adversaries are already working on fine tuning the open source ones for their benefits or making something even better. So just sit back and watch the fireworks.
China can train 70B and 400B LLM models using pre-sanction Nvidia GPUs.
The only reason they are loudly announcing "Hey our military is using YOUR open source models" is to trigger fearmongering and overzealous regulation in the West and make us less competitive in AI.
Who knows tomorrow we might be the enemy.