And isolating AGI from wireless networks of any form. And feeding learning data in that isolation. That should be in mutual agreement across all countries developing AGI. Because otherwise you set the world at risk.

Bad enough with localized language models in public. But these still have limited intelligence and autonomy. AGI is supposedly with limiters as well. But it is intelligent enough, and powerful enough to break those limiters. Otherwise you wouldn't be calling it a threat.

And they will contest that there are safety measures. But why would you call it a threat of autonomy then? So they will circle you around like this. But bottom line is that, it's resolved the moment you don't give it limbs.

And the responsibility befalls into human beings, of which, human beings can be monitored of their actions. AGI cannot be.