Combatting 'safety drift', more Meta-inspired layoffs and Europe's speech adversary
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.
I'm writing today's edition from the south of France, where it's Labour Day. I spent the morning milling through a nearby town, where locals marked the moment in 1890 when workers marched for the eight-hour day. It took another 29 years for it to become law, a reminder that change is slow and arduous.
Whether the AI transition will take quite that long is anyone’s guess, but if this week is anything to go by, Trust & Safety professionals and data labellers are on the front line of it (see: Platforms).
A big bienvenue to new free subscribers from Google, University of Berkeley, Hinge, The Home Office, The Quantum Hub, KonTerra Group and elsewhere. If you want the full-fat EiM experience, become a full member and get access to 450+ editions of Week in Review and T&S Insider.
I'm travelling next week so the next Week in Review will be on 15th May. The fantastic Georgia Iacovou (of Horrific/Terrific fame) will be guest editing Alice‘s Monday newsletter for the next month or so so be sure to look out for that.
Here's your Week in Review — BW
The tech industry is making measurable progress against online child sexual exploitation and abuse. In 2025, Tech Coalition member companies advanced 317 child safety mileposts—more than double the previous year.
These concrete improvements span prevention, detection, and response, reflecting how safety is being integrated earlier and more systematically into digital products and services.
As threats evolve, so does the industry’s response.
Policies
New and emerging internet policy and online speech regulation
Türkiye became the latest country to ban social media for teens, after legislation was passed to prevent under-15s accessing digital services including gaming platforms. According to Al Jazeera, the decision to go beyond social media comes after a a tragic shooting in southern Türkiye was reportedly carried out by a “bright but troubled teenager who spent a lot of time playing war games on his computer”. President Recep Erdogan also claimed in a speech that social media was “corrupting our children’s minds”.
From what I’ve read, the evidence is mixed about the effect of gaming on user’s aggressive behaviour and violence. If you’re closer to the evidence here — particularly from a T&S or research perspective — I’d love to hear from you.
Meanwhile UK government officials pre-empted the outcome of its ongoing under-16 social media ban consultation (EiM #333) after announcing that “age or functionality restrictions” would go ahead — but failed to say what that would look like or when it would be implemented.
Also in this section...
- The Hidden Debate Behind a €120 Million Fine (The Regulatory Review)
- If We Ban Social Media for Children, What Do We Offer Instead? (Digital Serendipities)
Alice Hunsberger
Products
Features, functionality and technology shaping online speech
Fine-tuning foundation models can introduce “safety drift” that weakens safeguards in unpredictable ways, according to a new joint report from the Center for Democracy and Technology and MIT researchers. “Out of Tune: Fine-Tuning Foundation Models Leads to Unpredictable Safety Drift” outlines how small adjustments of large AI lab models can have significant safety impact and challenges the assumption that base models remain durable after post-training and fine-tuning. The report strikes me as having potentially significant impact for how AI governance evolves and who is responsible for safety issues that arise. Very much worth a read.
Also in this section...
- Families sue OpenAI over failure to report Canada mass shooter’s behavior on ChatGPT (The Guardian)
- Real-time deepfakes are rewriting the rules of child safety (WEF)
💡 Become an individual member and get access to the whole EiM archive, including the full back catalogue of Alice Hunsberger's T&S Insider.
💸 Send a tip whenever you particularly enjoyed an edition or shared a link you read in EiM with a colleague or friend.
📎 Urge your employer to take out organisational access so your whole team can benefit from ongoing access to all parts of EiM!
Platforms
Social networks and the application of content guidelines
It continues to be a tough month for business processing outsourcers (BPOs) following the news that Meta is going all in on AI and reducing headcount across its business. Last week, it was staff at Sama who were affected (EiM #334) and this week “hundreds” of workers at Irish outsourcing firm Covelen were told their roles are at risk after cuts to Meta-related work, according to the Irish Independent. Wired also has more details.
Fight for the labour rights: It’s no wonder that researchers having been raising the alarm about the economic future of the average worker and AI labs have been releasing high-minded documents about their economic vision for society. Right now, it feels like the only winner from the “AI efficiency” drive is company shareholders.
However, a notable counterpoint came from a Chinese court this week, where a worker won a case for unlawful dismissal after a judge ruled that “organizational restructuring and reduced staffing needs” due to AI was not, on its own, sufficient justification for redundancy. It’s an early signal that the legal frameworks — in some jurisdictions at least — are catching up.
Ben WhitelawPeople
Those impacting the future of online safety and moderation
Sarah Rogers doesn’t have direct day-to-day impact on most T&S professionals, but her rhetoric feeds into a broader narrative that casts suspicion on online safety work as well as the people doing it.
As U.S. undersecretary of State for public diplomacy, the former attorney has become a prominent figure in Donald Trump’s administration and a key player in shaping how American free speech ideals are projected globally. That remit now includes leading the U.S. Agency for Global Media, which oversees federal-funded broadcaster, Voice of America.
In a good Politico profile, Rogers positions herself as a “forceful advocate for that spirit of the internet… where you can go to be free” — a framing that won't make her many friends in the EU. Not that she cares; the EU’s Digital Services Act is, in her view, “particularly offensive”, while the UK’s Online Safety Act is dismissed as “tyrannical and absurd”. Charming.
Posts of note
Handpicked posts that caught my eye this week
- “After months of hard work, access is open to our free, on-demand, e-learning course designed to build understanding about the world of video gaming, where harms can happen, and strategies to help adults and young people react, respond and build resilience.” - As someone whose gaming experience consisted solely of Crash Bandicoot and Pro Evo, I look forward to digging into a new resource from Christchurch Call’s Emily T.
- “Digital technologies have become one of the EU’s most regulated domains. Yet we rarely ask what this means for EU integration more broadly: is "digital" an exceptional case or the avant-garde of future EU policymaking?” - Policy nerds will love Matteo Nebbiai’s review of three recently published books on digital regulation.
- “With children's online safety top of the agenda, along with anxieties about digital surveillance at an all-time high, plus trust in technology companies at an all-time low, this topic often feels like it has too much 'heat' on it and not enough 'light’.” - I’m somehow reassured by having Dr Holly Powell-Jones in a room with other experts at a recent London workshop
Member discussion