How certain can we be, that AI Code Assistants or the companies behind them are "stealing" our proprietary code?

They are not stealing. They are merely expanding boundaries of the term "our code"

I raised this question last year in my YouTube video lol lmaooo...

guess security is still a concern 😕

if you can't prove they did it in court, along with you paying for the lawyers to actually go prove it, and the judge actually punishes them in some meaningful way from it that would actually encourage them not to do it

then they will take that into their liability calculation and it probably would be too little potential consequence for them to care about so they're doing it or gonna do it once they get the devs on plugging all those things in

They already stole literally everything once to train the models and got away with it. Why wouldn't they steal literally everything again?

@cuddlyogre yeah, I think the code stealing is done too, with all that code combined, it contains probably everything what we will write in the future as well. We'll, repos is one thing. But all those books, how could they get away with it. Name a very smol detail from a HP book and it knows. How can it deny having the sources. I think, they'll end up with a mixed dataset with no traces of authors and delete the real dataset.