I think this is too generous of an interpretation of OpenAI's and Anthropic's position. the 'slow down' mandate is not a real regulatory standard - policies are generally crafted based on good vs bad outcomes, not speed-of-development. the drive for this kind of specific, toothless regulation is done at the behest of large corporations looking at open-weight models eating their cake and realizing that any small player could now do so too if they had the compute backing them (places like Canopy Wave, for eg)

that it also serves as marketing material for just how capable their models are is just a cherry on top, I suppose