Spoiler; it's about content moderation rather than child protection [0], and it doesn't work [1]

[0] - https://arxiv.org/pdf/2505.10588 [1] - https://me.mashable.com/digital-culture/57507/ai-models-dont...

Tangential, do we need to "develop LLMs" with our content or is it a relic of when LLMs cant fit a novel into context?