Week in Review

Hello and welcome to Everything in Moderation's Week in Review, your need-to-know news and analysis about platform policy, content moderation and internet regulation. It's written by me, Ben Whitelaw and supported by paid members like you.

Welcome to new free subscribers from Nextdoor, Caleb, Respectify.ai, Google, Media24, Microsoft, eSafety Commission and elsewhere. Plus a big thanks to paying EiM members that support the newsletter and get access to EiM’s professional guides that were recently described by one happy member as “well worth the subscription“.

The Meta settlement is the big story of the week (see: Platforms) but it’s not the only US legal case worth your attention. Hit reply with your comments and queries; thanks for reading — BW

Policies

New and emerging internet policy and online speech regulation

A fascinating and divisive ruling coming out of the US where a federal judge has ruled that the First Amendment protects someone's right to privately possess AI-generated child sexual abuse material in their own home, provided it doesn’t depict a real child. The decision emerged from the prosecution of a Wisconsin man charged in 2024 with using an AI image generator to produce explicit images of children. The Washington Post and others have the story.

Legal experts quoted in the WashPo story are split on how much this actually changes anything: Stanford's Riana Pfefferkorn notes it doesn't protect producing, downloading or transporting such material, only pure private possession, while Boston College's Daniel Lyons thinks there's a real chance the Supreme Court takes up the case given the volume of synthetic material now overwhelming NCMEC's reporting pipeline. One to watch carefully.

Also in this section...

• Dutch regulator fines Uber $966 million for automating driver suspensions (Reuters)
• The Hidden Trade-Offs of Fighting Child Abuse Online (CACM)
• Parents and politicians love social media bans for kids. Experts worry they’re a shot in the dark (Politico)

Products

Features, functionality and technology shaping online speech

Alice — the AI security firm formerly known as ActiveFence — has raised $140 million in a round led by Apax Digital Funds, with Samsung, SentinelOne and a clutch of existing backers also participating. The company already counts Anthropic, Google and Cohere among its customers, and says it now works with eight of the ten leading AI model labs, testing them with malicious prompts and agentic tasks before release and monitoring inputs and outputs once they're live.

With the recent trend for "rogue agents", it is little surprise the company has had significant investor interest. It also validates its 2022 pivot from content moderation classifiers, which Alice (Hunsberger, that is) wrote about earlier this year.

A vendor rebranded with my name – here’s what it tells us about T&S tech

A T&S vendor just rebranded with the same name as this newsletter’s author — which is funny at first, until you consider what it reveals about the pace and direction of change in the industry

Everything in Moderation*Alice Hunsberger

Also in this section...

• Machines editing the rules that machines enforce (Zentropi)
• Our AI reality belongs to Huxley, not Orwell (Euractiv)

Platforms

Social networks and the application of content guidelines

You’ll almost certainly know by now that Meta has agreed to pay up to $18 billion to settle claims brought by dozens of US states that it knowingly harmed young users. The deal includes changes to how Facebook and Instagram work for teens, including a two-hour daily limit and blocks of features during the school day and at night. There's a lot to say and not much time in this week's Week in Review to say it so look out for next week’s T&S Insider for more.

Legal eyed: One outstanding question I have is that extent to which a settlement like this creates a legal precedent. Obviously, Meta has denied liability and there's no ruling on the merits for anyone else to cite . That said, Julia Smakman of the Ada Lovelace Institute noted how it gives EU regulators merely a useful baseline to work from while my Ctrl-Alt-Speech co-host Mike Masnick goes further in saying that it carries “the imprimatur of law” because of the push for YouTube and TikTok to adopt the same product features. I’m not a legal scholar and will let others advise on whether it stands up in court. But could it be that we, as an industry, need to be be careful not to read a verdict against one company, in one country, as a template for how the rest of the internet gets fixed.

A TikTok trend for filming cars speeding the wrong way down motorways has had tragic consequences after five teenagers died and others were critically injured in a recent crash. Such incidents are up year-on-year, despite the Chinese video platform claiming it removes activities that could cause significant physical harm and could be imitated,  and reporting by the Irish Times and the Journal.ie has highlighted that it’s a broader problem than the public — and perhaps the platform — realise. Either way, TikTok rejected an invitation to address a parliamentary committee on media, preferring to let the police and Coimisiún na Meán investigations play out.

Also in this section...

• OpenAI staff observed warning signs before AI agent hacking crusade caused global alarm (The Guardian) 
• Indian creators adopt coded language to evade platform moderation (Digi Watch) 
• YouTube, LinkedIn, and Substack Are Flagging AI-Generated Content. Brands Need a Public Policy (Inc.com)

People

Those impacting the future of online safety and moderation

Bill Gates isn’t well-known for his strong views on internet safety and, you could argue, has become more known for his views on philanthropy than technology. But I’d argue that a 6,000 word essay about AI from one of the original Big Tech titans is still worth noting. 

In it, he characterises the next decade of AI innovation and transformation as “one of the most turbulent times in human history” and makes the claim that it will be “either be the greatest equalizer ever invented, or the worst source of injustice”. The justification for that view is a little underwhelming but his mention of AI companions and discussion of labor displacement — not to mention bioterrorism and mass unemployment — makes it noteworthy for even the least engaged T&S worker.

What’s more interesting about the essays that it marks a shift in the way he thinks about AI from “super optimistic” in 2024 to a grounded optimism in 2026 with “deep concern about the AI harms we need to minimize”. Which reiterates something I’ve thought for a while: question anyone who says they are certain about the positives and drawbacks of AI. There are too many questions yet to be answered, to many unknowns, to many unexplored pitfalls, for anyone — T&S leader or otherwise — to claim certainty.