The other AI alignment problem, Discord takes a guess and new Ofcom probe
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 are a few spaces left for Monday's inaugural EiM Community Call with Alice and myself. This is a new way for EiM Members and readers to expand their network and learn from others. This month, we're talking about how to challenge misconceptions about Trust & Safety. Come along.
Welcome to new free subscribers from the BBC, Checkstep, Shopmy, Bureau ID, Fieldfisher, the Oversight Board and a host of Wikimedia folks, who must've caught wind of this week's T&S Insider. Have a read and then brush off your Wikipedia credentials and get stuck in.
Here's the major stories from the last seven days — BW
Policies
New and emerging internet policy and online speech regulation
Forget getting AI to share human values; the UN has a struggle on its hand to get member states to share a position on AI regulation right now. That was the takeaway from this week's General Assembly in New York, where AI safety was high on the agenda following the recent global panic (EiM #354). Here's what the key figures said and did:
- Donald Trump rejected calls for AI regulation, likening the dangers to the climate change "hoax", and asked that the technology be called "superintelligence" because "it is not fake. It's actually amazing." Hmm.
- Emmanuel Macron addressed the AI regulation question by calling for an open-source initiative that was independent of the US and China and did not "depend on somebody else".
- Ahead of meeting Trump for the first time, UK Prime Minister Andy Burnham said Britain was capable of being an "honest broker" in the global discussion on AI governance, citing the AI Safety Institute as proof of its credentials. )Talking of honesty, can we talk about the AI minister's audio setup?)
- Despite Burnham's big talk, the UK isn't one of the signatories of a 22-country declaration which calls to keep AI "under human direction, oversight and control". This looks like the most likely candidate to create an international AI oversight regime but lacks US and Chinese backing so far.
- Officials from both countries met on the sidelines of the General Assembly but, according to reports, didn't get far ahead of Xi Jinping's first US visit in a decade. The Chinese leader departs Washington today, so we should hear more over the weekend on any AI safety agreements. The Council on Foreign Relations has the issues to look out for.
The Philippines is dialling up its online safety efforts on multiple fronts following a recent school shooting, which a government official blamed on children's "access to deep dark web, Satanism, nihilism, anti-LGBT, chat groups encouraging you to commit crimes and to be glorified".
On Monday, the Senate unanimously passed the Sagip Bata Act, which criminalises grooming and luring, requires platforms to prevent and disrupt abuse, and would replace the country's existing anti-online child sexual abuse law. The next day, the country's privacy regulator issued show-cause orders to Meta, Reddit, Roblox and Discord, saying they were not complying with the country's data privacy laws. By the time this hits your inbox, I'm expecting other announcements.
In the UK, a little over a year after it trumpeted bringing pornography websites to heel, Ofcom announced on Wednesday that it is investigating Aylo, owner of Pornhub and other adult sites. The regulator is concerned that its age assurance, which relies on signals from Apple, did not undergo "sufficient due diligence and testing" to meet the highly effective age assurance bar. All eyes on this one.
That's not the only thing in Ofcom's in-tray: Meta is challenging the categorisation of WhatsApp and Instagram under the Online Safety Act claiming the most stringent Category 1 obligations aren't designed for messaging apps — which feels like a thin attempt to evade legally mandated duties. The Financial Times and others have the story.
Also in this section...
- The Overlooked Issue Undercutting EU’s Out-of-Court Dispute System (Tech Policy Press)
- Why AI companies can't be trusted to self-regulate (Nature)
- Beyond the fine: What the TikTok case reveals about LGPD enforcement (IAPP)
- Bootstrapping Frontier AI Governance by Mutualizing Risk (Lawfare)
In this week's Ctrl-Alt-Speech Podcast....
This week, join Mike and I to hear whether the AI apocalypse will be sexy (he suspects not but I'm still hopeful) and how the Tinder Swindler is making his comeback in, yep, dating apps.
Lots more serious analysis of the week's online speech and internet regulation news. Patreon supporters get extended and early episodes.
Products
Features, functionality and technology shaping online speech