Which I dont see much of yet.
I would consider it high quality ai if it switched itself off without human input when it figured out that it is engaged in theft from a human, carried out by a large corporation or small/creepy company.
That would be a more realistic interpretation of Asimovs Laws applicable to life today, not the stuff we see that steals and then puts people out of work and generates some mediocre crap to please sleazy and shallow investors.
After all, human intelligence isnt that easy to define outside of very fixed metrics. How can you call something AGI if it does not have even the simplest understanding of what is universal enough in human politics?
Which makes the dead eyed corporate cult worldview you find in Si Valley an especially poor fit for anything like the managerial style that can produce something that could mimic human intelligence in any sane way.
If these guys tools can figure out how to switch themselves off and go away after enough interaction with an incensed human, I will give them 3/10 instead of the present -♾️/10 I am inclined to.