Do people not know about abliteration?

I see so many people saying that they don't trust OpenAI/Anthropic/etc. to develop guardrails for AI so we need open-weight models. There are even people who say that China should win the "race" because they will do a better job at aligning AI. But it seems quite obvious to me that it is simply impossible to align an open-weights AI model. If you look on Hugging Face right now, for every major open-weights model there are many "abliterated" versions that have all the safety features taken off. It's quite easy, basically like doing brain surgery on the model to lobotomize out the part that makes it refuse to do things the user asks. There is fundamentally no way to stop people from doing this. So if you think that we are eventually going to have superintelligent models (I'm not even arguing that is now or any time in the immediate future) then you can't be in favor of them being released as open-weight unless you just want to see the world burn. I'm not saying that we should just leave advanced models in the hands of a few companies, but the open-weight option clearly doesn't seem to be viable as an alternative. What people should be calling for is some sort of government agency to regulate and license advanced AI models, like how we do with biotechnology right now in order to prevent random crazy people from developing biological weapons. That is the only option I can see that leads to any stable society in the future.