Anti-terrorism 'accountability theatre', Meta's Muse problem and Jess Davies vs xAI
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.
There's been so much in the news recently about AI oversight that I asked Alice to revisit her February 2025 T&S Insider piece on why it's time to unite Trust & Safety and AI governance. The result is a must-read if you're considering how to set up a team to succeed in this messy, fast-moving moment.
Thanks to EiM readers who joined Monday's Community Call and reminder that you can still sign up to be notified about October's event. If you want priority access, become an EiM member before this Sunday and I'll also send an EiM pin badge in the post.
Welcome to new subscribers from all over the map — France, Canada, the UK, India, South Africa, France, the U.S, and elsewhere. Here are the stories I paid most attention to this week — BW
P.S. I'm looking for EiM readers to give feedback on its new logo and design - if you're interested in taking a look and sharing your thoughts, hit reply.
Webinar: AI, emerging harms, and online safety
Join us October 20 as Ruth Leemis, Senior Researcher at Thorn, shares what the latest findings from Thorn’s Youth Perspectives Report mean for trust and safety teams. The session will cover the growing role of AI, why platforms need to account for young people accessing age-restricted spaces, and opportunities to better protect young users at scale. Unpack what stood out to our researchers after surveying more than 1,000 young people.
Policies
New and emerging internet policy and online speech regulation
Brazilians go to the polls on Sunday to elect governors, senators and federal, state and district representatives and, crucially, a president. The last time they did so the result was disputed for weeks, led to Jair Bolsonaro supporters storming Congress, the Supreme Court and the presidential palace (EiM 273) and the former president being given a 27-year prison sentence. Let's hope this time goes more smoothly.
To mark the occasion, I invited Katie Harbath, Meta's former director of public policy for global elections, onto this week’s Ctrl-Alt-Speech podcast — out now for Patreon supporters — where we talked about some important online speech related threads:
- AI model outputs: Brazil's election rules ban AI platforms from ranking candidates or suggesting who to vote for, but the chatbots are still doing it. ITS Rio and federal prosecutors found that 20% of responses from the country's most-used AI tools recommended a candidate in its testing.
- Deepfake definition: The electoral court (TSE) has decided that synthetic content is banned only if it is realistic enough to pass as a real person and also counts as campaign material. It will be tested to its extreme over the next few weeks,
- Digital advertising: Meta looks like it will have to explain itself to officials after it briefly stopped Lula's campaign from running ads and wouldn't say which rule he had broken. Meanwhile, campaign group Ekō found 26 accounts — some run from Guatemala, Colombia and Peru — which spent $223,000 on election ads, including unlabelled deepfakes.
It feels like an age ago now, but over the weekend culture secretary Lisa Nandy confirmed that the UK's social media restrictions for under-16s will come into force in March 2027. However, her interview with The Guardian told us more about what's coming than any of the lines coming out of Labour party conference. She: