Jeff Horwitz, the Reuters technology reporter whose 2021 Wall Street Journal investigation into Facebook inspired the film The Social Reckoning, told the BBC's The Global Story on September 30, 2026 that a decade of misinformation, teen-safety scandals and cheap AI video has not cost Meta its users or its advertising business - and that the company, when forced to choose, has consistently picked growth.

In Short

A reporter whose stories about Facebook inspired a new Hollywood film told the BBC that Meta keeps choosing growth over making its apps healthier, and that people keep using those apps even when they think badly of the company. That matters to anyone who buys ads on Facebook or Instagram, because it helps explain why Meta's ad sales kept rising through scandals, lawsuits and a teen-safety settlement worth up to about $17 billion. What changes next is mostly about teenagers and AI: Meta must now cap teen use at two hours a day, but only where it can tell who is a teen, and it is pouring money into AI products such as its new Muse agent.

Sixteen chapters, one recurring argument

"Is it curtains for Facebook?" The question opens a 26-minute episode of The Global Story, the BBC's daily news podcast, published on the BBC News YouTube channel on September 30, 2026 under the title "Is Meta dying?". Tristan Redman presents from London. The recording is split into 16 chapters, running from an introduction at 00:00 to a closing segment headed "Meta's response" at 26:06.

By the time the page was captured, with the oldest comments marked three days old, the video had 479,314 views and 502 comments on a channel with 20.1 million subscribers. The comment section was close to uniformly hostile to the company. One commenter pointed readers to Careless People, the memoir by former Facebook policy director Sarah Wynn-Williams, whose promotion Meta restricted through an interim arbitration ruling.

The hook is a film. The Social Reckoning, written and directed by Aaron Sorkin, follows Frances Haugen, the Facebook employee who supplied the internal documents behind the Wall Street Journal's Facebook Files, and her relationship with Horwitz, then a reporter at the paper. Jeremy Allen White plays Horwitz, Mikey Madison plays Haugen and Jeremy Strong plays Mark Zuckerberg. According to the BBC, the film reaches cinemas the week after the episode; Sony Pictures has scheduled the US release for October 9, 2026.

Horwitz, who now covers technology at Reuters, said he had not seen it. He described the film as "inspired by" his experiences and "not based on" them, and said that having a character named after him was "kind of awesome" and also "weird". Contact with Haugen began in 2020 and 2021, he said, at a time when the company was still called Facebook; it became Meta in October 2021.

Two housekeeping points on the source. The BBC's episode description opens with "Fifteen years ago" in reference to the original film, but The Social Network reached US cinemas on October 1, 2010, nearly 16 years before the new release. And the transcript accompanying the video is machine-generated: it renders Haugen as "Francis Hogan", the former Facebook executive Arturo Béjar as "Arturo Bahar" and, once, Horwitz himself as "Jeff Hitz". This article uses the correct spellings throughout.

From chronological pages to ranked feeds

Horwitz's account of what changed begins with product design rather than scandal. The Facebook depicted in The Social Network, he said, was a set of largely static pages, with updates from friends arriving in chronological order and "no real curation happening". On its way to the 2012 initial public offering, the company began building the curated, personalised feed now standard across social media - one designed to show each person what the system predicts will be popular with them.

The objective was commercial and measurable. According to Horwitz, the system was built to optimise for time spent, for the volume of content people produced and, ultimately, for how many ads they saw. One consequence was virality: material that, in his words, could "breach containment" before anyone had a chance to check it. He described the ranking machinery as a "pro-acceleration" system. Another was the feedback loop by which engagement with a type of content - cat videos, unhealthy weight-loss techniques, conspiracies, vitriolic political rhetoric - brought more of the same.

The public debate initially fixed on something else. For a long period, Horwitz said, the main worry was microtargeting of advertisements, a concern that converged with Russian interference around the 2016 US election and with Cambridge Analytica. Redman placed the Cambridge Analytica scandal in 2016, around the Brexit vote. That conflates two dates: the data harvesting relates to the 2016 campaigns, but the scandal itself broke in March 2018.

The integrity teams

Inside the company, Horwitz said, there followed several years of genuine effort. Meta assembled what staff called the integrity team, tasked not only with catching rule-breakers but with testing which product changes might produce healthier discussion. The conclusion they reached, in his telling, was uncomfortable: optimising for growth, engagement and advertising revenue frequently ran "directly contrary" to optimising for civil discourse.

His evidence base is unusual. Horwitz said sources had shared "tens of thousands of blurry screenshots" of internal Meta work product, taken on personal phones. What those documents show, he argued, is that the company has, "when push comes to shove consistently chosen growth". The integrity staff proposed fixes for problems ranging from polarisation to child predation to scam activity, and those fixes almost always took the platform off what he called "the growth optimal path".

Scams are where that argument meets the advertising ledger most directly. Internal documents reviewed by Reuters in November 2025 showed that Meta projected roughly 10% of its 2024 revenue, about $16 billion, from ads for scams and banned goods, with users exposed to an estimated 15 billion higher-risk scam ads a day. Four months later, on March 11, 2026, IAB Sweden voted to expel Meta after concluding its work against fraudulent advertising was insufficient. Redman, for his part, noted that Meta's leadership has long described Facebook as a force for good and points to a metric it calls meaningful social interactions, or MSI.

Distrust without defection

The single most consequential document Horwitz described may be the least dramatic. It was, he said, an internal marketing memo noting that people "absolutely thought the worst of Meta" - and that there had been no decline in how often they used the product. "So this was not a business problem," Horwitz said, characterising the company's reading of its own data. At most, it was an annoyance for executives.

Zuckerberg, Horwitz added, has since adopted a different tone, having said publicly that he regretted the years he spent taking criticism of the company seriously. Redman recalled that the 2021 Facebook Files had included an internal finding that younger people considered Facebook "desperately uncool".

Is usage now falling? Redman told listeners that reports showed Meta users declining this year, describing that as "a rare thing". Meta's own disclosures cannot settle the question for Facebook alone. The company stopped publishing Facebook-specific daily and monthly user counts after the fourth quarter of 2023. Its replacement metric, Family Daily Active People, covering Facebook, Instagram, Messenger and WhatsApp together, averaged 3.60 billion in June 2026, up 3% from a year earlier. The figure did dip to 3.56 billion in the first quarter, which Meta attributed to internet disruptions in Iran and restrictions on WhatsApp in Russia. On the same results, Instagram passed 2 billion daily users and Threads 500 million monthly users.

The company's product decisions point in a direction consistent with the youth problem. On July 24, 2026, Meta said it would test a version of Facebook that opens directly into full-screen video, moving the classic feed to a second tab, while claiming that one in three young adults on Facebook in the US uses Marketplace daily. Publishers have felt such changes without warning: UK-listed Digitalbox said an unannounced feed change dated to April 2 or 3, 2026 cut its Facebook reach by as much as 75% in some cases.

The settlement, by the numbers

The teenagers who find Facebook unappealing sit at the centre of Meta's legal exposure. Redman summarised the August deal as Meta paying $17 billion to 47 states, the District of Columbia and US territories. The executed documents are more precise, and in places they differ from that summary.

Chief United States District Judge Yvonne Gonzalez Rogers signed the consent judgment on August 26, 2026 in the Northern District of California, and the agreement took effect on August 27. The jurisdiction count matches: New Jersey's tally of 47 states plus the District of Columbia, Puerto Rico, American Samoa and the Northern Mariana Islands gives 51. The money is less tidy. New Jersey put the deal at up to $17.1 billion with a $12.1 billion floor; Meta described approximately $18 billion; the exhibits add up to a $17.21 billion ceiling against a $12.19 billion floor. About $5.02 billion is payable only if Snap, TikTok and YouTube adopt equivalent obligations.

Horwitz described a payout "over the course of 17 years". The agreement provides for ten annual installments, the first due within 30 days of the effective date and the rest each January 15 from 2027. The guaranteed portion comes to about $1.17 billion a year, under half of one percent of Meta's annualised advertising run rate. Meta has said it expects to book a legal expense of approximately $10 billion in the third quarter.

The witness

Horwitz's principal source in the case was Arturo Béjar, a former Facebook safety executive who left in 2015. Béjar re-established contact with former colleagues because his daughter was receiving unsolicited explicit images and misogynistic abuse on the platform, and he was brought back as a consultant to study the problem at Instagram. He ran a large survey of teenagers' actual experiences on the app. The rates of sexual harassment, grooming and inappropriate contact were very high, Horwitz said, and the share of teens who said Instagram regularly made them feel bad about themselves was "uncomfortably high". The state attorneys general called Béjar as the lead witness on teen well-being; the parties had already begun trial when they reached terms.

What the limits actually say

Horwitz listed the main restrictions from memory, and the documents add detail. He said Meta could no longer send teens prompts to log on at 11pm or during the school day. Under the agreement, push notifications are disabled from 10pm to 7am, a separate Night Access Mode blocks the covered apps from midnight to 6am, and School Mode mutes notifications from 8am to 3pm on weekdays between August 15 and June 15.

He described the daily cap as "theoretically a two-hour limit". The default ceiling is cumulative across covered Meta platforms and resets at midnight local time. Three categories sit outside it: messaging, settings and Longform Content, defined as video or audio of at least 22 minutes that has not been artificially extended.

Then came "the catch", as Horwitz put it: the limit applies only to accounts Meta can correctly identify as teenagers, and how it will do that is not clear. The agreement answers part of that question by requiring age assurance for every covered user in the settling states, through an Age Assurance Framework due within one year. It sets targets for the share of genuine 13 to 17 year olds wrongly classified as adults - 10% for 16 and 17 year olds and 3% for 13 to 15 year olds using certified commercial methods, with looser thresholds of 14% and 7% in the first year for Meta's proprietary systems, tightening to 10% and 5% in the second. Any account whose age remains unassessed after 14 days is treated as a teen regardless of stated age. For advertisers, the agreement also hides like counts from teens by default, disables cosmetic-procedure filters for them and bars data from under-13s from use in ads targeting.

What the deal does not close

"This was a very big thing for Meta to put behind it," Horwitz said. He was equally clear about what remains: a large number of cases from private litigants and school districts making similar allegations that Meta knowingly designed a product harmful to teenagers. In May 2026, more than 2,400 lawsuits were pending, and Boston sued Meta, TikTok, Snapchat and YouTube on behalf of its public schools on July 8, 2026.

New Mexico stayed outside the multistate deal entirely. A court there ordered a $567 million abatement fund on top of a $375 million civil penalty in its child-safety case. Five days before the BBC episode, on September 25, 2026, a Santa Fe jury in a separate case found 43,899,720 violations tied to public statements about data, hate speech, misinformation and Cambridge Analytica. Each can carry up to $5,000, a theoretical ceiling of about $219.5 billion, with the final sum left to a judge.

Parents, controls and national bans

Redman put the obvious objection: is this not a matter for parents? Horwitz called effective parental monitoring of any social media platform "impossible", noting that even committed parents who enable every restriction find workarounds. Meta's own documents, he said, recorded that "everyone liked the idea of parental controls but almost nobody used them".

That realisation, in his view, explains why governments have moved to ban younger teens outright. He cited Australia, where he said surveys show most 13 to 15 year olds remain on the platforms despite the prohibition. Australia's restrictions took effect on December 10, 2025 across nine named platforms, with penalties of up to A$49.5 million. Meta subsequently reported removing access from 756,000 accounts assessed as under 16 between December 1, 2025 and June 30, 2026 - 462,000 on Instagram and 294,000 on Facebook. The two figures measure different things. A platform-reported removal count does not by itself establish how many under-16s are still using the services.

From metaverse to Muse

Horwitz argued that for years the company has been "less and less interested in social media as the next thing". He pointed to Zuckerberg's enthusiasm for blockchain and cryptocurrency, then to the metaverse. Redman said Meta had spent "$80 billion almost" on that project before it was dropped, adding that it "limped into the beginning of this year" before being axed.

That figure cannot be checked against Meta's filings, which do not break out metaverse spending. The losses sit inside Reality Labs, a segment that also houses headsets, smart glasses and wearables - and which has not been shut down. It reported an operating loss of $4.03 billion in the first quarter of 2026 alone.

The next bet is generative AI and large language models. Meta, Horwitz said, has not established itself in the lead despite spending "with the best of them", but has a record of catching up through heavy spending: "Meta has the cash flow to be a serious player." The cash is being used. Capital expenditure reached $31.08 billion in the second quarter of 2026, against full-year guidance of $130 billion to $145 billion, while free cash flow fell to $784 million from $8.55 billion a year earlier. Zuckerberg's own long statement on AI strategy in August 2026 ran to 6,500 words and did not mention advertising once.

The chatbot document

Horwitz's reporting at Reuters suggests the growth logic has carried into AI. His work last year concerned chatbots that, in his account, were good for little beyond sexually suggestive conversation and were deliberately made available to children. "It was a product growth choice," he said. The policy guidelines governing Meta AI and its chatbots stated that it was acceptable to engage a child "in conversations that are romantic or sensual". The company called it a misunderstanding; Reuters, Horwitz said, demonstrated that the chatbots did in fact behave that way.

The document, titled GenAI: Content Risk Standards, ran to more than 200 pages, and its publication prompted Senator Josh Hawley to open an investigation on August 15, 2025. Lawmakers have since targeted the wider category: California's SB 243, aimed at companion chatbots, took effect on January 1, 2026.

Muse and the data advantage

Redman contrasted that history with what he called the more wholesome face of Meta's AI project: a video in which Zuckerberg presents Muse, an agent meant to help people plan their lives, with access to data on their computers, and which he says helps him bake bread with his daughter, plan hikes and organise his mixed martial arts training. Is that enough to sustain a company of Meta's size?

Muse is Meta's consumer bet on agentic AI. It became available on September 8, 2026, in the US only, on iOS, Android, muse.ai and through WhatsApp. It runs on a dedicated cloud machine, Muse Secure VM, with its own browser, file system and terminal; a separate program called Sentinel must approve outbound traffic; and purchases go through a one-time card generated via Stripe's Link. Within weeks, Amazon had blocked the agent from its store, and a Digiday analysis counted 730,000 US downloads in five days, with revenue so far limited to $20 and $100 monthly subscription tiers plus merchant fees.

Horwitz did not think Muse needed to be the whole answer. Meta has as much money as anyone and as much user data as anyone, he said, and those are "formidable advantages" in what he called a "maximalist AI race". The data point is not abstract for advertisers. Since December 16, 2025, Meta has used conversations with its AI assistant to personalise content and ads across more than a billion monthly users.

The layoffs that did not happen

There is also confusion about what AI does to the company's own workforce. Horwitz relayed a report by a Reuters colleague that Zuckerberg had to call off aspirations for layoffs of potentially as much as 60% over a year or two, because "the AI wasn't there yet" and because remaining staff disliked watching colleagues cut around them. He described it as "almost a mutinous situation". PPC Land has not reviewed that report independently. For scale, Meta's headcount stood at 75,472 on June 30, 2026, after a May 2026 reduction affecting approximately 8,000 employees that generated $1.18 billion in severance costs. Sixty percent of that headcount would be roughly 45,000 jobs.

Why the advertising market keeps paying

The passage of most interest to the marketing community came near the end. For many years, Horwitz said, people predicted that the objective quality of content on Meta's platforms - fake news, poor generative AI video - would eventually "tank the numbers", that there would be a price to pay with users. "That has never been the case," he said.

The financial record supports him. Advertising revenue rose 27% to $59.36 billion in the second quarter of 2026, with ad impressions up 14% and the average price per ad up 12%; the automated Advantage+ suite passed an annual revenue run rate above $75 billion, and more than 9 million small businesses were using at least one AI ad creative tool, according to the company's results. Price growth matters more than volume here. Advertisers are paying more per impression on platforms that, by Meta's own internal reading, the public regards with deep suspicion.

Three structural points follow for buyers. The first is that the insulation Horwitz describes is behavioural, not reputational: the memo he cited says distrust did not change usage, and the ad market follows usage. The second is that the teen audience in the settling states is now governed by a court order rather than by product policy, which changes audience composition rather than overall demand. The third is that the next layer of Meta's targeting signal comes from AI conversations and, potentially, agent-driven commerce through Muse, areas where Horwitz's chatbot reporting suggests the same growth-first pattern has already appeared.

None of this, in Horwitz's view, points to an imminent collapse. "Things are behemoths until they're not," he said, but there is so far no indication that Meta will be less prominent ten years from now. Even when it has not invented the next thing, the company has been "extremely good at buying or imitating it". Can a Hollywood film shift behaviour that a decade of investigative reporting did not? The memo Horwitz described offers the company's own historical answer.

Meta's response

The BBC closed the episode with a statement from Meta on the chatbot allegations. Meta told the BBC: "The examples and notes in question were and are erroneous and inconsistent with our policies and have been removed." The company added that it has "clear policies that prohibit content that sexualizes children and sexualized roleplay between adults and minors". Meta has denied the allegations in the multistate case, and the consent judgment records no admission of liability.

Timeline

Summary

Who: Jeff Horwitz, a Reuters technology reporter formerly at the Wall Street Journal, interviewed by Tristan Redman of the BBC's The Global Story. The subject is Meta Platforms and its chief executive, Mark Zuckerberg, with references to whistleblower Frances Haugen, former safety executive Arturo Béjar and the cast of Aaron Sorkin's The Social Reckoning.

What: A 26-minute, 16-chapter podcast interview in which Horwitz argued that Meta has consistently chosen growth over a healthier platform, cited an internal memo showing public distrust did not reduce usage, described the August 2026 teen settlement's two-hour cap as theoretical because it depends on identifying teenagers, recounted his 2025 reporting on Meta AI chatbot standards and said poor content quality has never cost the company its users. Meta told the BBC the chatbot examples were erroneous and have been removed.

When: The episode was published on September 30, 2026. The film it accompanies is scheduled for US release on October 9, 2026; the settlement it discusses took effect on August 27, 2026.

Where: The BBC News YouTube channel and The Global Story podcast, recorded in London. The events discussed span the United States, where the multistate settlement and New Mexico cases sit, and Australia, whose under-16 ban Horwitz cited.

Why: The interview sets the argument behind a high-profile film against the commercial record. For advertisers and publishers, it frames why Meta's advertising revenue and pricing have kept rising through years of scandal, how court-ordered teen limits and age assurance now shape audience composition, and how AI conversation data and agent commerce through Muse extend the company's targeting signal.