Meta's $18B teen safety settlement has a hidden play against TikTok and YouTube. Plus: is training AI on copyrighted books actually legal?
Meta agreed to an $18 billion multi-state settlement over claims it designed Facebook and Instagram to be addictive and harmful to kids. The financial penalty is the headline, but the real story is buried in the terms: Meta is offering $5.3 billion to the 29 states in the suit (Oklahoma included) to pursue its competitors. Ashley breaks down why Meta's biggest fear was never the teen account restrictions, it was being the only platform that had them.
Then: is it legal to train AI models on copyrighted books? A judge ordered Anthropic to pay $1.5 billion to authors, but ruled the training itself was lawful. The problem was where the books came from. Daniel unpacks the fair use question, why copyright law hasn't been meaningfully updated since 1976, and what this precedent means for independent creators whose work was scraped without permission or payment.
AI news you should know about, in under 10 minutes.
In this episode:
(00:38) Meta's $18B settlement and the teen account restrictions
(01:28) The $5.3B incentive to go after TikTok and YouTube
(02:37) Beauty filter limits and what they actually cover
(03:52) Anthropic's $1.5B payout and the piracy distinction
(04:48) Why 1976 copyright law can't handle this
(06:53) Sora, scraped creator content, and getting paid zero
Links: Show notes include "If Meta's going down, it's taking TikTok and YouTube with it" (The Verge) and the full AI copyright piece.
Follow Daniel at @danielhillmedia on Instagram, Threads, and YouTube. Follow Ashley at @ashleyrcoffey89 on Instagram, and grab the free AI newsletter linked in her bio.
Ashley Coffey (00:04)
Hello and welcome back to AI news that you should know about for August 28th. I'm your host, Ashley Coffey
Daniel Hill (00:11)
And I'm Daniel Hill. There's a lot of AI news happening in the world. We give you the highlights in under 10 minutes. This week on the show, we'll be talking about Meta settling a $17.5 billion lawsuit, which is more complicated than you might think. We'll also talk about whether or not it's illegal to train AI models on copyrighted books. We'll be covering all of that on our show today, but first, a quick word from our sponsors.
Ashley Coffey (00:37)
Welcome back.
Ashley Coffey (00:38)
Let's dive in. Alright, Meta has agreed to a landmark, $18 billion multi-state settlement to resolve a massive lawsuit accusing the company of designing Facebook and Instagram to intentionally be addictive and harmful to children. Nobody's talking about the craziest part of this settlement though, and it's what exactly Meta wanted to get with their legal jujitsu and the
Ashley Coffey (01:01)
The key to why they settled on the table on top of the huge financial penalties are the restrictions to teen accounts. Now, the biggest fear wasn't the rules. Their biggest fear was being singled out as the only one with the rules. If Meta caps teens at two hours and they meet that two-hour cap, they're gonna go to TikTok or YouTube or anything else. And so Meta, what they're doing is they're dangling $5.3 billion to the states that were a part of this lawsuit, which are
Ashley Coffey (01:28)
29,
Ashley Coffey (01:29)
Oklahoma being one of them. And they're basically saying you know what, we're gonna offer you $5.3 billion if you hunt and catch our competitors. So Meta is now getting handcuffed and realizing they have restrictions and is trying to spin it as being an innovator and a trailblazer and incentivizing the government to go after its rivals. I had an ad pop up on my feed right after this news broke, and I laughed so hard when I saw it.
Ashley Coffey (01:53)
saw it because I knew exactly what they were doing here. And I'll go ahead and put that ad up on the screen now so you can take a look at it. But essentially what that ad says is an open letter to TikTok and YouTube to join us in supporting teens. What a way to spend that. Daniel, what are your thoughts?
Daniel Hill (02:10)
I'm
Daniel Hill (02:10)
actually really glad that you pointed out that angle because I hadn't really thought about it. Everything you said is right. If teens reach a limit on the time that they can spend on Facebook or Instagram, they are going to hop over to a very similar app that gives them very similar styles of content. However, it's clear that what Meta has done for teens is so incredibly damaging. And some of the other things that they've agreed to as part of this settlement are things that I think are badly needed. One of the things is that they're
Daniel Hill (02:37)
limiting more access to beauty filters. Extreme makeup filters are going to be restricted from teen accounts. Now that doesn't include effects that make users look like a fantasy character or an animal or other effects that distort people's appearance. It doesn't include filters that smooth skin or alter appearance in ways that can be achieved by ordinary makeup techniques. But I do think this is steps in the right direction, but meta's still trying to protect their pockets.
Daniel Hill (03:04)
It always comes down to this. This is such a small amount in their entire revenue amounts that they make on a regular basis. So interested to see whether or not they're going to get TikTok and YouTube to agree to this settlement in the same way and that same amount.
Ashley Coffey (03:18)
Absolutely. And it'll be interesting for us to all see this kind of play out because all this stuff is so fresh. The judge has not approved this settlement yet. It's highly unlikely that he would decline it because Meta's already going out and doing this PR press release. So I'm going to link an article in the show notes called If Meta's Going Down, it's taking TikTok and YouTube with it from the verge. That way, if you want to learn a little bit more about this, you can take a look at that.
Ashley Coffey (03:41)
Alright, we're gonna take a quick break, but when we come back, we'll be talking about is it legal to train AI models on copyrighted books? It's complicated.
Daniel Hill (03:52)
Welcome back, continuing with our next story. Is it legal to train AI models on copyrighted books? It's complicated. By way of background, last year there was a ruling in the United States where a judge ordered Anthropic to pay a $1.5 billion copyright settlement to a group of writers whose works were used to train Anthropics AI models. That seems like it was a moral victory for authors. However, the judge ruled that Anthropic's AI training was lawful.
Daniel Hill (04:20)
What was the problem in this specific case was that Anthropic was pirating these books from illegal online libraries. Interestingly, the judge compared the way an LLM ingests trillions of words to a writer's study of literature, which you can agree with or disagree. That's fine. I don't want to get too deep into that. I do want to talk about fair use, which is basically copyright law that allows for the use of copyrighted materials without permission.
Daniel Hill (04:48)
If you are going to either criticize them, parody them, use them for education or other means. That's the important thing to keep in mind. Okay, let's get into the discussion. This isn't news per se, but I think it's an important thing to talk about, which is that copyright law really hasn't been updated since 1976. Judges who are faced with legal court cases about this need to interpret guidelines from 50 years ago when none of this existed. The article that we'll link to in the show notes.
Daniel Hill (05:14)
Includes a quote from an attorney who works in IP. And he said, quote, they know that the AI model has been trained on so much stuff and the law has really not caught up. End quote. So the question here is, is this fair use? What AI companies are doing? Are AI companies taking written words and are they actually transforming them in a way that qualifies as fair use?
Daniel Hill (05:38)
This is really important because it's going to set the scene for how the AI industry shapes up over the next coming months and years. Ashley, what's your thought on this?
Ashley Coffey (05:47)
I had no idea copyright law hasn't been updated since 1976. Wow. I feel like it's absolutely due for an update. What I have a problem with
Ashley Coffey (05:56)
That's how all these AI companies were able to produce the data set that is the basis of all these models that we know and use right now. So if they were to say that, you know what, that should be protected, they would have to figure out how to pay all these people around the world for the content that we so freely gave to be trained. And we all know that that's never gonna happen. And they're just trying to get it in writing, is what they're trying to do. And it's really weird to kind of think about that happening now.
Daniel Hill (06:23)
But I agree with your exact sentiment. And that was one of the things that I wanted to discuss is if these companies are slated to make billions of dollars off of these copyrighted materials as a result of ingesting this information, they can come up with some agreed upon number. Let's say the price of a hundred copies of an author's book, whatever it is, whatever the agreed upon metric is. But these companies certainly have large enough pockets and they're building their businesses based on this information.
Daniel Hill (06:49)
They can compensate the authors fairly and they can certainly afford to do that.
Ashley Coffey (06:53)
I think about us as content creators because whenever they originally built Sora, they basically scraped the internet of all the freely available YouTube videos and Instagram videos, which we'd have a lot of those. So I sometimes think about it, you know what? All the stuff that we're using, I know my stuff was used to train it. How much money did I get? Zero dollars. How much money did you get? Zero dollars. It's
Daniel Hill (07:13)
Correct. Zero.
Ashley Coffey (07:15)
really weird and I feel like the precedent that's being set only favors larger corporations and not independent creators like us.
Daniel Hill (07:22)
Yes. And I think the model or the idea is, okay, well, we're taking advantage of the tools that in some cases are free, but they're not always going to be free. Once we get used to using the tools, the prices are going to go up. So I contributed my data without my permission. You used my data to train these models, but how does that benefit me in the long run? It doesn't. And I 100% agree with your point.
Ashley Coffey (07:44)
just feels like human labor being exploited. Again.
Daniel Hill (07:47)
Yes, we
Daniel Hill (07:48)
are working for free without having given permission in order to do so. Yes. That is it for our AI news you should know about this week. The show notes have links to everything we've covered. And if you want more content like this, please connect with us on social media. You can find me at Daniel Hill Media on Instagram, Threads, and YouTube.
Ashley Coffey (08:03)
And you can find me on Instagram at AshleyRcoffey89. And if you want the AI highlights without the twice-a-week commitment, I do have a free AI newsletter. You can find the link to sign up in my bio.
Daniel Hill (08:14)
Follow the show, send us any topic you would like us to discuss on here, and we'll see you next week for more AI news you should know about. Thanks for listening.