

they will most likely know how to use it properly and ethically
I’d argue that ethical use is not possible:
- Models are trained on stolen/misappropriated/misused data
- Training involves psychologically harmful work from ghost workers
- Those services runs on infrastructure that no one wants around, and wastefully contributes to climate change/global warming













For yt-dlp specifically, there is a stable backport that is fairly up-to-date: https://tracker.debian.org/pkg/yt-dlp
It doesn’t yet have yesterday’s version, but the previous version was made available 3 days after it was released.