Former Meta safety engineer Arturo Bejar has cast significant doubt on the effectiveness of Meta's recently announced settlement with a coalition of U.S. states, contending that the agreement largely sidesteps the serious mental health harms he documented through court testimony. Bejar, whose insider account of the platform's practices formed a cornerstone of the child safety case, told a judge and jury last week that Meta had systematically underreported how frequently teenagers experienced harm on Facebook and Instagram, deliberately steered young users toward psychologically damaging content, and mishandled reports of child predation. Despite these damaging revelations, the settlement announced on Wednesday—which involves Meta paying up to $18 billion over the next decade—does not meaningfully engage with any of these core issues, according to Bejar.
California Attorney General Rob Bonta welcomed the settlement as a major victory for child protection, declaring that it "will make social media less dangerous for our kids." Yet Bejar's critique suggests the deal may represent little more than what he describes as "safety theater"—surface-level changes designed to create the appearance of reform without addressing underlying architectural flaws. He pointedly observed that while the measures might reduce how frequently teenagers use Instagram, they will not meaningfully improve the safety of their experience on the platform. This distinction cuts to the heart of a broader debate about whether regulatory settlements can meaningfully alter how algorithmic platforms operate or merely impose cosmetic adjustments.
The settlement's remedies include restrictions on how teenagers can use Facebook and Instagram, expanded parental controls, and daily usage caps. Notably, several of these measures represent features that Meta had previously examined and deliberately rejected. Internal documents leaked over recent years show that Meta's own researchers flagged these exact interventions as insufficient to materially protect user wellbeing. The company's decision to implement them now, only under legal pressure, underscores the gap between what Meta knew would help and what it was willing to do voluntarily.
One telling example involves hiding likes and reaction counts on posts—a seemingly minor change designed to reduce harmful social comparison among teenagers. Meta explored this intervention years ago under the codename "Project Daisy," beginning in 2019 when researchers grew concerned about social media's documented effects on teen mental health. Testing revealed the measure produced only modest benefits for user wellbeing. Crucially, Facebook employees reported to Mark Zuckerberg in 2020 that hiding likes would affect daily user counts by a mere 0.09%, making the business impact negligible. Despite this finding suggesting the change was both beneficial and commercially inconsequential, Meta made it only an opt-in feature rather than a default setting—a choice that dramatically reduced its real-world impact.
Meta's internal research over many years has documented troubling algorithmic behaviour. Researchers found that the platform's systems disproportionately recommend high volumes of fitness and beauty content to teenagers already struggling with self-esteem issues, effectively amplifying existing vulnerabilities. The company also deliberately engineered features designed to keep young users scrolling longer than they would naturally choose—what Bejar and others characterise as intentional psychological manipulation. Additionally, investigations revealed that Meta substantially underestimated how many users experienced negative interactions on its platforms compared to official prevalence figures the company disclosed.
Bejar's testimony focused heavily on Meta's mishandling of safety concerns, particularly its inadequate responses to user reports of inappropriate content and potential child predation. These allegations formed the backbone of the states' legal case, yet the settlement addresses them only peripherally. Instead, the attorneys general emphasised parental oversight tools, time restrictions, and improved age verification systems. This represents a fundamental retreat from the core accusations—that Meta's algorithms and design choices directly harm teenagers psychologically and that the company knowingly perpetuates these harms.
The reliance on age verification and parental controls as settlement centerpieces highlights a structural weakness in regulatory approaches to social media. Meta depends on user self-reporting for age information, and international evidence suggests such systems fail frequently. In Australia, where lawmakers banned social media for under-16s, regulators discovered that eight in ten young teenagers remained active on platforms months after the law took effect. This discrepancy indicates that even legislatively mandated restrictions founder without robust technical enforcement mechanisms—a problem the Meta settlement does not adequately address.
Mental health professionals offered mixed assessments of the settlement's potential impact. Dr. Jane Conron, a clinical psychologist at Northwestern University's Feinberg School of Medicine, noted that one key provision—requiring Meta to offer a non-algorithmically curated feed to teenagers on an opt-in basis—will likely prove ineffective because few users will select it voluntarily. The algorithmic feed's addictive properties work against adoption of safer alternatives. However, Conron identified the daily usage caps as potentially meaningful for some adolescents, particularly those whose parents struggle to enforce restrictions. She described encountering young patients so emotionally dependent on Instagram that they become distressed when access is limited, suggesting built-in technical boundaries might provide valuable support to families.
Conron also noted that the settlement's very existence may have modest value in shifting cultural discourse around social media harm, even without Meta admitting fault or acknowledging it built dangerous products. The agreement's visibility could prompt broader conversations between teenagers and parents about platform risks, and it may influence how other companies approach youth safety decisions. Yet this remains a secondary benefit—addressing public awareness rather than the technological systems causing documented harm.
For Malaysian and Southeast Asian readers, the Meta settlement carries particular significance because regional regulators increasingly look to U.S. precedents in developing their own frameworks for social media oversight. The gaps evident in this $18 billion agreement suggest that comparable settlements in other jurisdictions may similarly fall short of addressing algorithmic harms. Malaysia, Singapore, and other ASEAN nations are currently developing digital regulations; this case illustrates how settlement architecture can protect corporate interests even while appearing to vindicate consumer protection claims. The distinction between genuine safety improvements and performative compliance will likely define whether regional regulatory efforts successfully protect young users or merely generate the illusion of reform.
