[jkestner]

Look, they’re only keeping those videos to train their model to identify CSAM and definitely not accidentally generate it.