I wonder, and this will date me a bit, but statistically how many separate AI implementations would be required to nvidia-chip into existence before the inciden

I wonder, and this will date me a bit, but statistically how many separate AI implementations would be required to nvidia-chip into existence before the incidental exploitation of a single WOPR-class device which happens to have a backdoor accessible by a teenager with public-school-district-level hacking skills.