The Algorithm Against Our Children

In March 2026, after a nearly seven-week trial in Santa Fe, New Mexico, a jury returned a verdict that should have broken open every boardroom in Silicon Valley. The jury found that Meta — the company that owns Instagram, Facebook, and WhatsApp — had knowingly harmed children's mental health, concealed what it knew about child sexual exploitation on its platforms, and engaged in "unconscionable" trade practices that unfairly exploited the vulnerabilities of young people. The penalty: $375 million. Juror Linda Payton, 38, explained what guided the jury's judgment. With a maximum $5,000 per violation at stake, she said she thought each child was worth the maximum amount. (Source: NPR, "New Mexico jury says Meta harms children's mental health and safety, violating state law," March 24, 2026, https://www.npr.org/ 2026/03/24/g-s1-115019/new-mexico-meta-children-mental-health; AP/US News, March 24, 2026, https://www.usnews.com/news/technology/articles/2026-03-24/new-mexico-jury-finds-meta-viola ted-consumer-protection-law-at-trial-about-child-safety)

The trial examined a substantial body of Meta's own internal documents and research — evidence that the company's own engineers and scientists had conducted studies on the impact of its platforms on young users and had not acted on what they found. The jury heard from Meta executives, platform engineers, former employees who had become whistleblowers, psychiatric experts, tech safety consultants, and public school educators who described classrooms disrupted by sextortion schemes targeting children. It heard about Meta's documented failure to enforce its own ban on users under the age of 13, and about the role its algorithms play in prioritizing sensational and harmful content — including content about teen suicide.

At the core of the case was a charge that New Mexico Attorney General Raúl Torrez pressed with moral clarity: Meta designed its platforms to addict young users. It knew what was happening. It chose profit. "We know the output is meant to be engagement and time spent for kids," prosecution attorney Linda Singer told jurors. "That choice that Meta made has profound negative impacts on kids."

More than 40 state attorneys general across the country have filed similar lawsuits against Meta. Thousands of families are in litigation. The coalition of parents who have lost children to harms caused by social media — the group calling itself ParentsSOS — called the New Mexico verdict "a watershed moment." "We parents who have experienced the unimaginable — the death of a child because of social media harms — applaud this rare and momentous milestone," the group said in a statement. (Source: PBS NewsHour, "What's next in social media legal battles after a New Mexico jury finds Meta platforms harm children," March 25, 2026, https://www.pbs.org/ newshour/nation/whats-next-in-social-media-legal-battles-after-a-new-mexic o-jury-finds-meta-platforms-harm-children)

A Letter That Changed a Country

A twelve-year-old girl in Australia named Charlotte O'Brien took her own life because of bullying that was enabled and amplified by social media. Her mother, Kelly O'Brien, wrote a personal letter to Prime Minister Anthony Albanese. He read it. It moved him. And it helped move a government.

Charlotte's story was not an isolated tragedy. It was the story that Australia's news organizations — and a campaign called Let Them Be Kids, backed by more than 54,000 petition signatures from parents, educators, and child safety advocates — used to make the scale of the harm visible to the people with the power to act. An event at the United Nations General Assembly in September 2025 featured a mother speaking about her daughter's suicide as "death by bullying... enabled by social media." The speech drew support from world leaders from Greece, Fiji, Tonga, and European Commission President Ursula von der Leyen.

On November 29, 2024, the Australian Parliament passed the Online Safety Amendment (Social Media Minimum Age) Act. It came into full effect on December 10, 2025. Australia became the first country in the world to enforce a nationwide ban preventing children under the age of 16 from holding accounts on major social media platforms — including Facebook, Instagram, TikTok, YouTube, Snapchat, Reddit, X, Threads, Twitch, and Kick. Platforms that fail to take "reasonable steps" to enforce the age restriction face fines of up to $49.5 million Australian dollars. (Source: Wikipedia, Online Safety Amendment (Social Media Minimum Age) Act 2024, https://en.wikipedia.org/wiki/Online_Safety_Amendment; eSafety Commissioner, "Social media age restrictions," March 2026, https://www.esafety.gov.au/about-us/industry-regulation/social-media-age-restrictions; Time, "What to Know About Australia's Social Media Ban for Kids Under 16," December 10, 2025, https://time.com/7339762/australia-youth-social-media-ban-under-16-snapchat-meta-tiktok/)

"This is the day when Australian families are taking back power from these big tech companies," Prime Minister Albanese said when the law took effect. "They are asserting the right of kids to be kids."

Jonathan Haidt, social psychologist and author of The Anxious Generation — a book that documented with systematic rigor the relationship between the mass adoption of smartphones, social media, and the mental health crisis now consuming an entire generation — welcomed Australia's action. "The world is rooting for your success," he wrote, "and many other nations will follow."

They are already following. As of February 2026, France, the United Kingdom, Germany, Italy, Greece, Spain, Denmark, Malaysia, and New Zealand are all considering or actively advancing similar restrictions for children under 16. The European Parliament passed a non-binding resolution advocating a minimum age of 16, and has proposed banning addictive features like infinite scrolling and autoplay for minors entirely. (Source: CNBC, "Australia is trying to enforce the first teen social media ban. Governments worldwide are watching," December 10, 2025, https://www.cnbc.com/2025/12/10/australia-16-year-old-teens-ban-social-media-policy-law-ig-tikt ok-fb-reddit-youtube-snapchat.html; Australia's Social Media Ban: A Test for Global Digital Governance, BISI, January 2026, https://bisi.org.uk/reports/australias-world-first-social-media-ban-a-test-for-global-digital-governa nce)

The Privacy Trap Hidden Inside the Protection

I want to say something honest about the age restriction approach — because I believe the governments moving in this direction are doing the right thing for the right reasons, and I also believe that the implementation they are being pushed toward carries its own serious dangers.

Age verification, as it is currently being discussed and deployed, requires social media platforms to confirm a user's age before allowing account creation. The methods under consideration include government ID checks, biometric facial scanning, behavioral inference from browsing history, and AI-based age estimation from physical appearance. Every single one of these methods requires the collection of sensitive personal data. Some require biometric data. Some require linking social media accounts to government identity documents. Some involve building a digital profile of a user's behavior, demographic characteristics, and physical appearance — and retaining it.

The same companies I have documented throughout this book as serial violators of user privacy — companies that added "terrorist" to Palestinian users' bios by algorithm, that scanned 6,167 browser extensions without user knowledge, that shared data with foreign governments at a compliance rate of 94 percent — these are the companies now being positioned as the enforcers of children's age verification. The data they will collect to enforce the restriction is exactly the kind of data they have historically misused, monetized, and handed over to governments on request.

Mozilla, the non-profit organization that builds the Firefox browser and has spent decades defending user privacy, raised this concern directly in response to the Australian law. Age estimation systems based on biometric data, Mozilla noted, often underperform for people with darker skin tones and non-binary facial features. Behavioral inference systems built to estimate age based on browsing history create detailed profiles of every user. Age verification via government ID creates a linkage between social media identity and legal identity that has never before existed at scale. (Source: Mozilla Foundation, "Australia's Social Media Ban: Why Age Limits Won't Fix What Is Wrong With Online Platforms," December 19, 2025, https://blog.mozilla.org/netpolicy/2025/12/19/australias-social-media-ban-why-age-limits-wont-fix -what-is-wrong-with-online-platforms/)

Let me be direct about what this means. The companies whose platforms are harming children cannot be the companies trusted to verify the identities of those children in order to protect them from their own platforms. That is not protection. That is an expansion of the surveillance infrastructure under the guise of safety — the same move Big Tech has made with security, with content moderation, with anti-terrorism measures, every time the public has demanded accountability.

The protection of children from the harms of social media is a legitimate and urgent moral priority. The mechanism of that protection must not create a new and larger harm in its place. Any age verification framework must be independent of the platforms, privacy-preserving by design, subject to public audit, and built on the principle that compliance with a child protection law does not entitle a company to build a biometric database of every adult who uses its service.

The Algorithm That Rewards Rage — and the Children Who Pay for It

Let me now turn to something that is less discussed than the explicit harms of social media to children, but which I believe is equally consequential for the generation growing up inside these

platforms: the deliberate design of recommendation algorithms that reward the most extreme, divisive, and inflammatory content — and the financial incentives that keep that design in place.

In 2025, Sky News conducted a month-long experiment on X — the platform formerly known as Twitter, now owned by Elon Musk. The study created accounts with different political leanings and tracked what content was algorithmically recommended. The finding was unambiguous: X "exhibits a significant algorithmic bias toward promoting far-right and extreme content." The Wall Street Journal and the Washington Post conducted their own separate analyses. Both reached the same conclusion: X now amplifies right-leaning and extreme political content over other perspectives. (Source: Social Media Today, "Research Shows X Amplifies Conservative Political Views," February 19, 2026, https://www.socialmediatoday.com/news/x-formerly-twitter-amplifies-conservative-political-persp ectives-report/812652/)

A 2025 academic study published in Science used a browser extension methodology to alter users' X feeds during the 2024 U.S. presidential election — reducing exposure to content expressing anti-democratic attitudes and partisan hostility. The study found this single change shifted users' political feelings toward their opponents by more than two points on a 100-point scale — a shift that would normally take approximately three years to occur organically in the general population. The researchers described this as causal evidence that X's algorithm — its recommendation system, its financial incentive structure — is actively contributing to political polarization. (Source: The Conversation, "A few weeks of X's algorithm can make you more right-wing — and it doesn't wear off quickly," February 18, 2026, https://theconversation.com/a-few-weeks-of-xs-algorithm-can-make-you-more-right-wing-and-it- doesnt-wear-off-quickly-276153)

This is not an accident. It is an engineering decision. The algorithms on X, TikTok, Facebook, and YouTube are not neutral filters. They are optimization machines, designed to maximize engagement — and the content that drives engagement most reliably is the content that provokes the strongest emotional reactions: outrage, fear, contempt, tribal identity, moral panic. That is what keeps people scrolling. That is what keeps the session length high. That is what makes the advertising inventory valuable. And that is what is being served, disproportionately, to teenagers who are still forming their understanding of the world, their sense of identity, their relationship with people who are different from them.

The financial incentive structure compounds this. Platforms financially reward the creators who generate the most engagement — through monetization programs that pay per view, per watch hour, per share. A creator who produces content that provokes rage is not just getting more views. They are getting paid more for producing that content. The algorithm selects for the most inflammatory content, the most polarizing framing, the most extreme positions — and then writes a check to whoever produced them. The far-right content farms that now operate at industrial scale on X and TikTok are not producing what they produce because they believe in it. They are producing it because the algorithm has made it the most profitable content category available to a creator without significant production resources.

What this means for a fifteen-year-old spending four hours a day on TikTok or X — seeing that content, being financially excluded from the social spaces where their peers organize — is a daily immersion in a curated environment that makes the world seem angrier, more divided,

more dangerous, and more certain than it actually is. That is not an accident either. It is the product.

Silencing Palestine, Amplifying Extremism: The Algorithm's Double Standard

I documented in Chapter Nine the specific mechanisms by which Meta, TikTok, X, YouTube, and LinkedIn systematically suppressed Palestinian voices during the Gaza genocide — deleting posts, suspending accounts, complying with Israeli government takedown requests at rates as high as 94 percent, and erasing 700 videos documenting potential war crimes at the direction of the U.S. government.

I want to return to that record here, in the context of what these platforms have simultaneously allowed. While Palestinian journalists had their accounts suspended, their posts deleted, and their reach algorithmically throttled, the same platforms left anti-Palestinian hate speech online at a rate of 96 percent, according to the Centre for Countering Digital Hate. The 7amleh digital rights organization documented over 19,000 cases of hate speech and inciting content in Hebrew on X in the weeks following October 7, 2023 — content that remained online without consequence. (Source: Business and Human Rights Centre, "Report Exposes X's Failure to Remove 96% of Hate Speech Posts," November 2023, https://www.business-humanrights.org/en/ latest-news/report-exposes-xs-failure-to-remove-96-of -hate-speech-posts-amid-israel-palestine-conflict/)

The pattern is consistent across platforms and consistent over time. The algorithm suppresses the voices of the occupied and amplifies the language of the occupier. It silences grief and rewards rage. It removes documentation of atrocity and recommends the rhetoric that justifies it. And it does this not because of an ideology embedded in the software — but because of the financial architecture that the software is designed to optimize. Content that generates outrage gets more engagement. More engagement generates more advertising revenue. The atrocities are deleted because someone with government connections asked for them to be deleted. The rage is amplified because rage is what the business model runs on.

The same algorithm that suppresses the Palestinian child's testimony keeps the extremist content creator in the recommendation queue — because the extremist content creator generates more engagement, and engagement is money.

Our children are being raised inside this architecture. They are learning what the world is from platforms that are financially incentivized to show them the most extreme version of it. They are learning what counts as a human being worth grieving from platforms that delete some suffering and amplify other suffering based on which deletion and which amplification produces the most profitable outcome.

And when the consequences of that architecture become visible — when the clinical studies document the mental health damage, when the suicides happen, when the New Mexico jury finds that Meta engaged in unconscionable practices that exploited the vulnerabilities of children — the companies issue statements, update their community guidelines, and return to the same practice at the same scale in the next news cycle.

What Regulation Must Actually Require

The age restriction laws moving through parliaments from Canberra to Copenhagen to Brussels are a genuine and necessary beginning. They are not sufficient.

Protecting children from social media harms requires going beyond restricting who can hold an account. It requires addressing the algorithmic and financial architecture that makes those platforms harmful in the first place — because a sixteen-year-old is no less vulnerable to an algorithm optimized for outrage than a fourteen-year-old, and an adult who spends four hours a day in a recommendation system designed to maximize engagement and emotional reaction is being harmed too. The age restriction addresses the symptom — underage users — without touching the disease — the design.

What is actually needed, alongside the age restriction, is mandatory algorithmic transparency — the legal requirement that platforms disclose, to independent auditors with binding authority, precisely how their recommendation systems work and what they optimize for. Not a quarterly transparency report written by the company's PR department. An independent, binding audit with the power to mandate changes to algorithmic design when the evidence shows those systems are causing harm.

What is needed is liability for algorithmic amplification of harmful content. The Section 230 shield that protects technology companies from liability for content posted by users should not extend to content that the company's own recommendation algorithm amplified, promoted, and financially rewarded. When a platform pays a creator to produce content that drives a teenager toward self-harm, the platform is not a passive host. It is a participant. The law should treat it as one.

What is needed is a global ban on the algorithmic and financial reward structures that incentivize the most extreme content. The business model that converts engagement into revenue — and that therefore selects for the content that generates the most powerful emotional reactions — is the engine of both political radicalization and the mental health crisis in our children. You cannot regulate your way out of the consequences of that engine without addressing the engine itself.

And what is needed is the recognition, at the level of binding international law, that the platforms that operate across borders cannot be governed by the standards of the country most willing to look the other way. Australia moved first. Dozens of countries are following. The pressure on the European Union to enforce the Digital Services Act to its full legal capacity — holding platforms liable for algorithmic harm with the same urgency with which it holds them liable for privacy violations — is real, growing, and long overdue.

Charlotte O'Brien was twelve years old. The New Mexico jury found that each child harmed was worth the maximum penalty available under the law. So do I.

--- The harm I have documented in Chapter Eleven — to children, to their mental health, to their developing understanding of the world — is harm caused by platforms whose most dangerous capabilities are already visible and regulated. Or trying to be regulated. What Chapter Twelve describes is a category of AI capability that is not yet regulated, that the companies building it acknowledge could be catastrophic, and that is proliferating faster than any governance framework can keep up with. The Code Bomb is not a metaphor. It is a technical description of what Anthropic has already built and what every major AI laboratory is racing to match.