Meta Platforms' landmark $18 billion settlement with a coalition of American states announced this week has drawn sharp criticism from the whistleblower whose evidence proved instrumental in building the child safety case. Former Meta safety engineer Arturo Bejar contends that despite the considerable financial commitment and promised remedies, the agreement sidesteps the core problems he testified about regarding how the platform deliberately harmed young users.
Bejar's testimony to a judge and jury outlined damaging practices that Meta allegedly pursued systematically. He described how the company deliberately understated how frequently teenagers suffered harm on Facebook and Instagram, engineered the feed algorithm to steer young people toward content designed to degrade their self-perception, and demonstrated negligence in responding to reports of child predators operating on the platform. These allegations formed the crux of the states' legal action against the technology giant.
California Attorney General Rob Bonta hailed the settlement as transformative, stating it would meaningfully reduce dangers to young users. Yet Bejar's post-announcement assessment paints a starkly different picture. In his view, the agreement essentially formalizes what he calls "safety theater"—superficial gestures that create an illusion of protective action without substantively improving user safety. He argued that while teenagers might spend marginally less time on Instagram, the platform itself would remain fundamentally unsafe in its core operations.
Meta characterised the agreement as an extension of its existing commitment to supporting teen wellbeing and empowering parents. The company framed the settlement as representing genuine, enforceable transformation coupled with enhanced transparency. Company officials expressed gratitude for Bejar's contributions, even as they dispute the scope of necessary change required beyond what legislatures and industry peers could accomplish.
The settlement's structure reveals a troubling pattern when examined against Meta's own historical decision-making. Leaked internal documents have shown that several remedies now forming the settlement's centrepiece were previously considered by Meta and deliberately rejected. Most notably, Meta agreed to remove visibility of like and reaction counts on posts, ostensibly to combat "negative social comparison" among teenagers competing for peer validation through engagement metrics. However, Meta's own testing programme, codenamed "Project Daisy" in 2019, demonstrated that hiding likes produced only minimal effects on user wellbeing. Internal communications from 2020 explicitly stated that engineers observed no meaningful improvements in overall wellbeing measures when the feature was tested.
The business calculation further undermined the measure's potential impact. Meta's testing predicted that implementing the like-hiding feature universally would reduce daily active users by merely 0.09 percent—a negligible figure that explained Meta's reluctance to adopt it as mandatory policy. Instead, the company offered it only as an optional feature users could select, essentially burying it from broader impact. That this same feature now anchors the settlement suggests regulators extracted commitments Meta had already determined would pose minimal practical consequences for either user harm or revenue generation.
Meta's internal research has long flagged algorithmic systems that systematically amplify problematic content to vulnerable populations. The company's own scientists documented how recommendation systems deliver intensified flows of fitness and beauty content to teenagers already struggling with self-esteem, effectively worsening existing psychological vulnerabilities. Additional research highlighted intentional product design choices that psychologically manipulate young users into extended engagement far beyond what they would naturally choose. These algorithmic and design problems represent systematic rather than incidental harms embedded in Meta's business model.
Bejar's work at Meta specifically documented that teenagers experienced negative outcomes on the platforms at rates substantially exceeding Meta's official public prevalence statements. This discrepancy between internal reality and external claims proved central to the states' legal theory. Notably, the settlement addresses these algorithmic and moderation failures only peripherally. Instead, the remedies emphasise expanded parental oversight tools, daily time limits on app usage, and improved systems to identify minors falsely claiming to be adults—measures that treat symptoms rather than the underlying platform mechanics generating harm.
The challenge of age verification itself underscores the settlement's structural limitations. Meta and peer platforms rely fundamentally on user self-reporting for age classification, creating inherent enforcement weaknesses. Australia's experience proving instructive: when authorities banned social media for users under 16, subsequent monitoring discovered that eight in 10 young teenagers remained active on these platforms. This suggests that even with regulatory force behind age restrictions, technological and behavioural workarounds enable continued access. Meta's commitment to identify underage users and impose restrictions faces similar practical obstacles.
Expert assessment among mental health professionals remains divided on the settlement's real-world utility for adolescents. Dr. Jane Conron, a clinical psychologist at Northwestern University's Feinberg School of Medicine, emphasized that the settlement's core algorithmic protection—offering teenagers an opt-in non-algorithmic feed—would likely remain unused due to the friction created by making it a voluntary choice. Conron noted that the platform's intentional design psychology makes opting out of the algorithmic feed feel like abandoning social connection itself, rendering the choice largely theoretical.
However, Conron identified potential genuine benefit in the settlement's requirement that Meta implement mandatory daily usage caps for teenage users on Instagram and Facebook. She described clinical cases where young patients experience such severe emotional distress when parents attempt to limit screen time that they respond with tears and psychological withdrawal. An automated platform-enforced usage limit could bypass these manufactured emotional dependencies by externally constraining access. For certain vulnerable users already struggling with compulsive use patterns, this mechanical restraint might provide relief they struggle to impose independently.
While the settlement carries significance simply through its existence and the official acknowledgement of harm it represents, this significance remains primarily symbolic rather than structural. Conron noted that the agreement's very presence—without requiring Meta to admit fault or concede that it built products with inherent destructive properties—may shift cultural conversations about social media's role in adolescent development. As awareness spreads about the degree of concern warranting such a substantial financial settlement, parents, educators, and teenagers themselves may recalibrate expectations about platform safety.
The settlement's ultimate effectiveness for Southeast Asian users, particularly in Malaysia where social media penetration among teenagers ranks exceptionally high, depends largely on whether Meta implements the provisions identically across global markets or defaults to regional variations. Meta's history suggests customised approaches by geography, meaning protective measures adopted in the United States may face weakened implementation elsewhere. Malaysian regulators and parents should anticipate that additional local advocacy may prove necessary to ensure equitable protection for young Southeast Asian users.
