I think you're misunderstanding what I'm saying. I don't think its technically feasible for current models to be open source, because there is no source code to open. Yes, there is a harness that runs the model, but the vast, vast amount of instructions are contained in the model weights, which are akin to a compiled binary.
If we make large strides in interpretability we may have something resembling source code, but we're certainly not there yet. I don't think the solution to that problem should be to change the definition of open source and pretend the problem has been solved.
If everyone open sources their AI code, Meta can snatch the bits that help them without much fear of helping their direct competitors.