The Youth Lab

Exhibit 313, an internal Facebook study from November 2018 titled "Long Term Retention: The Young Ones Are The Best Ones and Other Learnings.” A smoking gun? Not by itself. But nonetheless revealing. Already in 2016, a leaked company email described Instagram’s overall company goal in terms of “teen time spent”. Therefore, by November 2018, Meta had quantified something its strategy had already been signalling for years: the younger someone joined, the more likely they were to stick around.

It was just one in a long line of shocking exposes in a trial that was only getting going when Meta decided to settle - crucially, not on the courthouse steps. But one week in, when they could see the writing on the Wall Post - cutting short a six-to-eight week proceeding that could have put substantially more of these shocking revelations around the company’s internal decision-making into the public record.

Here are the facts:

WHY THE CASE EXISTED
The states alleged that Meta deliberately designed Facebook and Instagram to “entice and retain” young users, misled the public about risks to children, and violated consumer-protection laws. They also alleged violations of the federal Children's Online Privacy Protection Act (COPPA) through collection and use of data from children under 13 without proper parental consent. Meta denied wrongdoing.

WHO BROUGHT IT
The federal trial included consumer-protection claims from California, Colorado, Kentucky and New Jersey, plus COPPA claims brought by 29 states. The eventual settlement went substantially wider: 47 states, D.C., Puerto Rico, American Samoa and the Northern Mariana Islands, with a separate Texas component.

WHAT META WAS FACING
The four states whose consumer-protection claims were being tried were expected to seek close to $200 billion in civil penalties. Meta had argued that its theoretical statutory exposure across the broader litigation could reach $1.4 trillion.

WHAT IT SETTLED FOR
Meta says the total is up to approximately $18 billion over ten years. Roughly $12.7 billion is guaranteed; around another $5 billion is conditional on competitors including TikTok and YouTube accepting comparable restrictions and settlements. The multistate court filing itself puts that component at $16.68 billion, with Texas adding another $1 billion.

THIS IS ABOUT MORE THAN JUST MONEY:

THE GOOD - META HAS TO CHANGE!

1. The settlement changes the product, not just Meta's bank balance. Under-18s in participating U.S. jurisdictions will have a default two-hour daily limit across Facebook and Instagram, removable only with parental permission. There are also midnight-to-6am and school-hours notification restrictions.

2. Some of the mechanics associated with compulsive use are being directly constrained. Teens get 15-minute usage prompts, hidden like counts, controls over autoplay and the ability to make a non-personalised feed their default; parents can require that non-personalised experience. Cosmetic-surgery and extreme-makeup filters are also restricted.

3. Age assurance becomes a real obligation rather than merely asking users their birthday. Meta has committed to stronger technology to identify accounts likely belonging to under-13s, including looking across multiple accounts. The agreement requires ongoing compliance around age assurance.

4. An independent auditor will test Meta's compliance annually for five years, while a new independent research foundation will receive consented user data to study teen wellbeing - significant as access to platform data has long been one of the barriers to independent scrutiny.

5. It could establish a de facto industry standard. If TikTok and YouTube adopt the framework, Meta's protections become substantially tougher: one hour per app rather than two hours across Meta's apps, and a 10pm–7am curfew instead of midnight–6am.

6. State attorneys general have demonstrated that product design itself can become the subject of consumer-protection litigation. That potentially changes what product teams have to consider before shipping engagement features.

SO FAR, SO GOOD, RIGHT?

THE BAD - THE MOST UNCOMFORTABLE PART OF THE STORY IS WHAT META ALREADY KNEW

1. Meta had literally quantified the value of getting users young. Youth engagement was a measured business variable not an accidental consequence.

2. Another Meta internal document read “Teens are hooked despite how it makes them feel. Instagram is addictive.” That wording was presented by the states during opening arguments; Meta disputes the states' characterisation of its internal research.

3. Instagram introduced Take a Break, their flagship wellbeing intervention in 2021. Adam Mosseri (the human face of Instagram) testified that at one point only 1.8% of teens were using it. Meta publicly highlighted that more than 90% of people who switched the feature on kept it on. It did not publicly highlight the 1.8% adoption figure.

(Meta eventually made the feature default for Teen Accounts in 2024 after California and Colorado had filed lawsuits in 2023.)

4. Reuters reported testimony from Instagram design director Francesco Fogu concerning internal data indicating teens encountered more harmful content than adults; some safety data were subsequently removed from certain presentations. Meta denied suppressing research.

5. The trial had been expected to run six to eight weeks. It stopped in its second week. Zuckerberg had been expected to testify. Thousands of pages of internal material and weeks of testimony therefore never received the public airing they might have received through a completed trial.

ARE YOU READY FOR THIS?

THE UGLY - WHAT IF $18BN IS ACTUALLY A GOOD DEAL FOR META?

1. Meta's share price rose. Reuters reports that Meta closed about 1% higher following the settlement. Investors effectively concluded that removing the litigation risk was worth more than the cost of the settlement. “Historic punishment” indeed.

2. Meta generated $200.97bn of revenue and $60.46bn of net income in 2025. Even the maximum $18bn settlement is equivalent to less than four months of one year's Meta profit. That is an extraordinary penalty in absolute terms. Relative to the earning power of the company paying it, not so much. This is not Big Tech's Big Tobacco moment.

3. The settlement does not dismantle Meta's recommendation engine. The underlying mechanism that turns attention into revenue remains largely intact.

4. Meta is not required to make a non-algorithmic feed the universal default. Critics including former Meta safety engineer Arturo Béjar argue that this leaves the fundamental recommendation architecture largely untouched. “limiting time on a harmful product isn't necessarily the same thing as making the product safe.”

5. In the most generous view possible one might argue that Meta is helping create an industry-wide floor so teens can't simply migrate to the least-regulated platform.A slightly more cynical take is that Meta has managed to make part of its own punishment contingent on imposing similar costs and restrictions on competitors.

6. A teenager in California just acquired protections that teenagers using exactly the same Instagram app in Cairo, Caracas, or Cork do not have. Why should child safety depend on the postcode of the attorney general prepared to sue Meta?

7. A completed judgment potentially creates precedent in a way a negotiated settlement does not. The trial ended. No verdict. Zuckerberg never took the stand. Meta admits no wrongdoing.

This may be the biggest settlement Big Tech has ever seen. Whether it becomes Big Tech's Big Tobacco moment depends on what happens after the cheque clears.

SO WHAT COULD HAPPEN NEXT

The good: Oakland becomes Big Tech’s Big Tobacco moment: the point at which “growth at all costs” finally acquires a cost large enough to change what gets built.

The likely: The guardrails get higher, the lawyers arrive earlier and teenagers get better protections but the trillion-dollar attention economy learns to adapt faster than regulators can constrain it.

The ugly: $18bn buys absolution without admission, Wall Street celebrates, the algorithms keep learning and the industry concludes that harming first and settling later is still a perfectly viable business model.

BRAND TAKEOUTS

This has been a long 52INSIGHTS so let’s keep this short!

1. Design and build like your internal research will one day be read out loud in court.

2. If your best customers can't easily stop using your product, ask whether that's a KPI or a liability.