7 min read

Four questions for the next child safety settlement

Meta and TikTok have both settled child safety cases brought by US states. After reading Daphne Keller’s analysis of the agreements, here are my worries — and some questions for next time around

I'm Alice Hunsberger. Trust & Safety Insider is my weekly rundown on the topics, industry trends and workplace strategies that trust and safety professionals need to know about to do their job.

While most people have moved on from the Meta/TikTok settlements, now is actually the time to work out what they say, and what we should seek in similar future negotiations. Daphne Keller’s close reading of both agreements got me there.

Drop me a line with whatever you're working on at the moment and remember to apply to join our upcoming EiM Community Call — the final one of 2027 — at the end of the month. Here we go! — Alice


IN partnership with Resolver Trust & Safety on training a CSAM classifierS
CTA Image

How do you train a CSAM classifier on data nobody on the team is allowed to see?

The Roke team behind Vigil AI explain how they built it in a secure, air-gapped lab using the UK’s Child Abuse Image Database, how they protect victims’ privacy, and why they won’t open-source it: released weights would help criminals find material and evade detection.

Resolver’s Athena brings the classifier to platforms, with no imagery retained.

FIND OUT MORE

Will the Big Tech settlements actually make kids safer?

If you’re a regular reader of EiM, you’ll know that back in August, Meta settled its child safety litigation with US state attorneys general (#352) and a month later, TikTok settled with Alabama (#357). 

Both were widely heralded as a turning point for youth safety online but they didn’t have the effect that many thought; shockingly, Meta’s stock price went up you read that right after the announcement while other states have since turned on those that settled for accepting “peanuts”. How are we meant to think about what happened? 

My five takeaways from the Meta settlement
Most headlines call it a win for teenage internet users. But I don't think that Meta setting the safety features for every major platform constitutes a success for anyone

Here, I’m very grateful to Daphne Keller's close read of both settlements, which was published by Tech Policy Press last week. Keller prepared open, searchable and annotated copies of both  Meta and TikTok agreements and a comparison table tracking the differences between them. They are a phenomenal resource and anyone who wants the full picture should start there.

After reading Keller’s analysis, I’m (once again?) left worried about the lack of evidence for the changes mandated in the Meta and TikTok cases in the name of child safety. 

I realise there’s little that can be done in these specific cases but I wanted to have a way of thinking through future agreements, when they inevitably present themselves. If you’re negotiating or scrutinising the next one, here are the four problems I see in these two, which I’ve turned into questions to ask at the end.

1. They count minutes, not what fills them

Both settlements treat time as the main driver of harm. This builds on the idea — present in CDC research and elsewhere — that teens who have four or more hours of daily screen time are almost three times more likely to have had anxiety symptoms and depression symptoms in the past two weeks than teens with less. 

As we know, though, correlation is not causation and an anxious or depressed teen may be more likely to spend time in front of a screen, rather than getting out and doing fun activities. The best longitudinal research points to a web of causes, with gender and existing mental health problems as the most consistent predictors, and bullying, sleep, and feeling connected at school all mattering. 

On TikTok, the settlement still allows for kids to watch videos at night as long as they arrive via messages or search. As Keller puts it, what gets blocked at night is feed functionality, not screen time.

2. Every teen gets the same rules

Universal limits assume social media affects every teenager the same way. It doesn't.

What we do have is a lack of firm scientific consensus:

  • The Surgeon General's advisory, published back in 2023, laid out challenges to social media use but also highlighted a whole bunch of benefits of using social media, including self-expression and connection for LGBTQ people and those with disabilities. 
  • The health advisory from the American Psychological Association (APA) also lays out that using social media is not inherently beneficial or harmful to young people, and that it may be psychologically beneficial for those in mental health crises or members of marginalized groups.
  • Save the Children's 2025 “Click, Scroll, Connect” report found that emerging harms like algorithmic bias, deepfakes and AI-generated abuse material disproportionately affect girls, LGBTQIA+ youth, children with disabilities and children from minority backgrounds. 

A two-hour cap — as outlined in the Meta settlement — doesn't address those harms. 

This one is personal for me. I was a queer teenager in the US South around 2000. I felt alienated, and I was desperately looking for community. I found it online, among punks, activists, and other queer kids, and it got me through some really dark times. Sure, I never had an endlessly scrolling feed but the value of online community was always very clear. We shouldn’t deny that from others. 

Get access to the rest of this edition of EiM and 200+ others by becoming a paying member