Meta Platforms, Inc. is confronting a transformation in the political economy of digital attention. The company’s principal advantages—behavioral data at immense scale, algorithmic recommendation, targeted advertising, automated moderation and extensive youth reach—are increasingly being recast as sources of legal, regulatory, safety and reputational exposure. The most consequential developments are concentrated in August 2026, supported by claims dating from late July and a small number carrying later-dated metadata. The strongest corroboration concerns the New Mexico youth-safety order, Italy’s deepfake-disclosure ruling, Australia’s under-16 regime, the UK’s digitally manipulated-image reports and the expanding litigation over allegedly addictive product design. Other claims are single-source observations and should be treated as indicators rather than settled facts.
The central investment question is therefore no longer whether Meta can attract engagement. It is whether the company can preserve the economic returns of engagement while demonstrating that its products do not deliberately exploit minors, obscure data practices or amplify emotionally inflammatory content. Regulation is moving from the realm of voluntary safeguards into the product itself: age assurance, notification controls, moderation explanations, human review, synthetic-media labeling, privacy governance and advertiser accountability are becoming operational requirements, as well as potential foundations for damages claims.
Key Insights
Youth safety has become a structural litigation and product-design issue
The most durable theme is coordinated scrutiny of youth exposure and platform architecture. A U.S. appeals court reportedly allowed approximately 2,400 lawsuits to proceed over allegedly addictive design 106. Other accounts describe more than 3,000 cases 86,95, as well as thousands of claims against Meta, Google, TikTok and Snap 23. The wider litigation includes state attorneys general, families and school districts: more than 1,000 school districts are seeking compensation for societal costs associated with youth mental-health crises 101, while more than 40 states are reportedly pursuing related claims 101. Separately, 29 states allege that Meta engineered Facebook and Instagram to be addictive to children 36; litigation involving Meta and those states is corroborated by two sources 99.
The legal pressure is not confined to pleadings. Meta faces a federal youth-safety proceeding in Oakland, where jury selection began before the scheduled trial 17, and Mark Zuckerberg is expected to testify personally 97. School-district settlements and verdicts indicate that the issue is moving toward financial exposure 94,100,108. Four states have separately accused Meta of intentionally designing Facebook and Instagram to be addictive 20,30. Families of four teenagers have filed wrongful-death-related claims 108, while another filing names Meta, TikTok, Snap and YouTube in allegations linking platform use to teen suicides 77,108.
The legal theories increasingly target product architecture rather than third-party speech. Plaintiffs allege that repeated-engagement features, the collection of children’s data and design choices intended to maximize time spent caused measurable harm 62,71. In New Mexico, the judge identified autoplay, infinite scroll, like counts and algorithmic recommendations as features that may collectively facilitate addictive or problematic adolescent behavior 63. Recommendation systems have likewise been cited in addiction lawsuits involving adolescents 100. The claims span children’s privacy, consumer protection, products liability and failure-to-warn theories 100.
The evidentiary foundation, however, is uneven. The New Mexico court accepted a causal relationship based on one witness despite the absence of scientific studies demonstrating causality 63. Research cited in the cluster also suggests possible reverse causation: children with pre-existing mental-health needs may use social media more intensively 63. Legal momentum is consequently more certain than ultimate liability.
Section 230 remains a significant defense. It generally shields platforms from liability for third-party content 2,21,59,107 and protects good-faith moderation and publication decisions 100,107. Yet plaintiffs are attempting to characterize Meta’s conduct as defective product design, inadequate warnings, privacy violations or platform-created behavior rather than the simple publication of user content 62,100. This distinction materially increases tail risk because it could narrow the usefulness of Section 230 without requiring its wholesale repeal.
Regulation is shifting from voluntary safeguards to verifiable age assurance
New Mexico provides the clearest example of judicial intervention in product design. The court limited under-18 Facebook and Instagram accounts to 90 hours per month 24,68,88, required notification blackouts 24,31,32 and generally disabled push notifications from 10:00 p.m. to 7:00 a.m. 68, as well as on school days from 8:00 a.m. to 3:00 p.m. 68. The order requires banners and informational screens explaining safety features and tools for inappropriate content 108, age proof for users Meta estimates are under 13 108, and hidden Like counts for New Mexico minors 92. The framework applies across Facebook, Instagram, WhatsApp and Threads 68.
There is nevertheless a jurisdictional complication: one claim indicates that Facebook’s low adolescent penetration in New Mexico could limit the basis for restricting end-to-end encryption 74. The court also reasoned that adolescent restrictions should be removed once a user is verified as over 18 63.
The New Mexico Attorney General plans legislation mandating age verification and a broader consumer-protection overhaul 90, arguing that legislation is necessary to prevent recurrence rather than merely punish past conduct 90. Age-verification laws have bipartisan support across Republican- and Democratic-led jurisdictions 90, while comparable federal proposals have been introduced by both parties 90. Public sentiment reinforces this direction: 66% of Americans favor mandatory verification for users under 16, and 85% believe social media is addictive for children 91. Support for increased oversight or age verification reaches 62% among Republicans and 71% among Democrats 91.
Australia represents the most significant international precedent. Its national under-16 social-media law is corroborated by three sources 1,13, is described as a world-first ban 16 and formally entered into force in December 15. Covered platforms must prevent under-16 users from creating or maintaining accounts 110, detect newly created accounts 110 and prevent previously identified minors from bypassing restrictions by changing their declared age 110. They must also demonstrate continuous reasonable efforts through recurring governance and testing 110. Existing users may face additional verification, usage limits or termination 110.
The Australian regime does not require formal age verification as the sole compliance method 110, but platforms are expected to use comprehensive age assurance rather than self-declared ages 110. Enforcement lies with the Australian eSafety Commissioner 110. Reported maximum penalties differ—up to A$49.5 million 110 and A$99 million 104—and the discrepancy should be resolved before exposure is quantified. The rules are intended to reduce harmful content and cyberbullying 110, but they also impose direct governance and access-control costs 16.
The execution burden is inseparable from a contradiction at the heart of digital regulation: stronger verification requires more information about users, while privacy authorities demand data minimization. Age checks introduce verification friction 110, require additional personal information 110 and may create unnecessary privacy exposure for children 110. Opponents warn that verification could expand state surveillance 76, while the systems themselves can be circumvented 71. WhatsApp’s self-reporting approach is considered unreliable 57, and additional verification may generate both false positives and false negatives 57.
Indian parental-consent rules add further uncertainty around age classification, adult-account validation, parental verification and identity checks 57. WhatsApp must prove a parent-child relationship while minimizing personally identifiable-information collection 57. Models in which governments or banks digitally sign an age threshold without disclosing identity to the platform could reduce the privacy burden 73, but they would require interoperable standards that do not yet appear established.
Meta has already removed 756,000 Australian accounts believed to belong to under-16 users, including 462,000 Instagram accounts 106. Compliance can therefore reduce the addressable youth user base and potentially engagement, even as it protects the company from larger penalties. The episode also exposes the tension between WhatsApp’s end-to-end encryption and demands for safety visibility: a court found that encryption’s privacy benefits outweighed adolescent risks 63, while automated scam monitoring could conflict with the platform’s privacy commitments 14.
Data monetization and profiling remain both commercial foundation and regulatory fault line
Meta’s business model converts behavioral data into targeting and behavioral influence. Algorithmic systems aggregate clicks, likes, locations and linguistic patterns into individual and group psychographic profiles 102. Adaptive targeting can incorporate purchase history, inferred personality, emotional states and biometric signals 102. Platforms collect likes, shares, searches, communications, browsing and engagement patterns 40, while surveillance-capitalism models convert such data into the capacity for behavioral modification 103.
Users frequently lack awareness of the implicit exchange of personal data for free access 40, lose control over the contexts in which voluntarily disclosed data is used 50 and lack sufficient understanding of collection, retention, consolidation, reuse and third-party sharing to provide meaningful consent 40. Opaque consumer profiling is specifically identified as a Google data-practice risk 82, but the regulatory logic applies directly to Meta.
The mobile advertising ecosystem extends this exposure beyond Meta-owned interfaces. Advertising SDKs can automatically transmit location data to advertising companies 52 and data brokers 51, sometimes after an operating-system permission is granted without the user understanding downstream flows 84. EFF reporting identifies automatic location-data collection and transmission 84, potential privacy compromise absent informed consent 52 and a mismatch between app-level permissions and actual SDK data flows 84. Developers often adopt default SDK settings without independently reviewing what is collected or how it is used 84, while financial incentives encourage continued location sharing 85.
EFF has called for restrictions on transfers to advertisers and brokers 85, legal accountability for SDK creators 84 and stronger mobile-advertising regulation 84, including restrictions on behavioral advertising itself 84. Meta should therefore expect increasing scrutiny of pixels, SDKs, real-time bidding and cross-context identity resolution, rather than scrutiny confined to visible consumer products.
The litigation trend is already visible in CIPA and GDPR matters. Nearly 4,000 lawsuits have targeted digital tracking and communication tools under CIPA 41, including claims involving Meta Pixel, Google Analytics, session replay and chat tools 41. Although these technologies are standard web infrastructure, they create compliance and litigation risk 41. In Europe, the CJEU held in Fashion ID that a site embedding a social button could become a joint controller 79, and in Wirtschaftsakademie that a fan-page operator could share responsibility for Facebook processing 79. A Dutch lawsuit alleges that MoPub real-time bidding violated GDPR and seeks EUR250–EUR2,500 for approximately 11 million users 81. These precedents increase the importance of data-flow mapping, consent propagation and contractual allocation of liability throughout Meta’s advertising ecosystem.
Opt-out governance presents another operational weakness. Companies may need to propagate opt-out flags for targeted advertising, data sales and profiling across all systems under Vermont privacy law 80. Failure to transmit user choices in real time creates pause-ad compliance risk 109. Browser-based consent could reduce intrusive pop-ups and user complaints 42, but questions remain over how consent would be updated or revoked 42.
The cluster also reports an Apple App Store study finding inaccurate privacy labels and no evidence of systematic verification 79, complaints over Apple’s IDFA practices 79 and 75% of tested iOS apps combining sensitive-data access with tracking-domain connections 79. These findings are not direct evidence against Meta, but they support a broader movement toward enforcement across the full data supply chain.
Automated moderation and recommendation expose the trust-versus-scale contradiction
Meta’s automated systems are indispensable to scale and a source of error. WhatsApp experienced widespread account-access disruptions as complaints multiplied 11, demonstrating that automated enforcement can disable legitimate accounts 10. Reported consequences include false positives, reduced access reliability, dissatisfaction, reputational damage and erosion of trust 9. Public reaction reflected empathy for affected users and distrust of automated enforcement 10. WhatsApp’s dependence on algorithmic enforcement 10 and the need for stronger human review and accountability 10 reveal the limits of fully automated safety systems. Citing the company’s Terms of Service in account-review notices 10 did not resolve the underlying transparency problem.
The same contradiction appears in content moderation. Facebook combines automated detection, manual review algorithms and policy-based penalties 58, yet prohibited content can evade controls 58. At the same time, nearly all cases reviewed by EFF involved content or accounts that did not violate stated platform rules 93. EFF found that most censored material was factual or educational rather than drug-selling content 93. Content concerning medications such as mifepristone can be categorized as prohibited drug activity, producing removal, account restrictions, de-ranking or shadowbanning 93.
Human-rights organizations report disproportionate enforcement effects on Palestinians, activists, women, LGBT people and non-English content 100. Amnesty concluded that Facebook amplification of anti-Rohingya hate speech and misinformation contributed to an enabling environment for the 2017 atrocities in Myanmar 100. Instagram is also facing criticism over Nazi-related content and manipulated images that trivialize the Holocaust 35.
Recommendation engines are optimized to detect reactions, not factual accuracy 98. False narratives on X reportedly receive 50,000–200,000 likes and can reach the top of feeds 98, while opaque recommendation systems can accelerate misinformation contagion 103. AI-generated misinformation across Facebook, WhatsApp, Instagram and other networks reportedly contributed to migration flows toward Ceuta, although the causal connection remains unverified 33,106. The EU has responded with a communication tool and requests for stronger anti-disinformation measures, surveillance and verification 7,19, while Meta and TikTok have implemented fact-checking related to Ceuta 22.
The policy dilemma is therefore not simple censorship versus freedom. More aggressive intervention may reduce misinformation and legal risk, but over-removal can reinforce allegations of political bias and suppress legitimate information. Meta has attempted to address transparency through source-code and user-control initiatives. X released the source code for its For You timeline and tools to assess shadowbanning 106, while Meta’s recommendation data properties differ fundamentally from typical LLM workloads 105. Investors should watch whether Meta can provide meaningful explainability without exposing proprietary ranking systems or facilitating gaming.
Facebook recommendations help creators find target groups 67. Creator networks use leaders, multimedia assets and strategic advice to direct engagement toward profitable groups 67. The same infrastructure can support legitimate distribution or industrialized engagement farming; its social value cannot be separated from the class interests embedded in the monetization system.
Engagement optimization may create adverse-content monetization liability
The cluster contains several single-source allegations that Meta funds or monetizes rage-bait and controversial content. ABC News Verify reported that Facebook directly funded controversial Australian creators, including a white nationalist and an anti-vaxxer, to produce rage-bait 28. The relevant pages at times appeared inconsistent with Facebook’s monetization policies 64. The Noticer, banned by X for hateful-profile violations, reportedly continued receiving Facebook monetization payments 65,66. Rage-bait is designed to provoke anger and increase comments, shares and engagement 58. A reported Groningen operation allegedly generated revenue through polarizing Facebook content 27, and ABC separately published an investigation into Facebook paying controversial creators 26.
These claims do not have the same corroboration as the youth litigation or Italy’s regulatory ruling and should be treated as reputational and control-system indicators rather than established misconduct. They nevertheless expose a direct contradiction: Meta’s engagement-based incentives may reward precisely the content whose social costs generate regulatory attention.
The same incentive structure appears in anti-data-center content. Engagement farming around such material was reportedly operated anonymously from Bangladesh 70, while anti-data-center messaging was popular across major social networks and creators did not perform materially worse on Instagram than elsewhere 70. Meta’s recommendation system can suggest related groups after creators publish 67, accelerating both legitimate civic discussion and coordinated manipulation.
Youth-led activism creates civic value but also heightened harassment and technology-facilitated violence risks for women and marginalized groups 71. HateAid reports a multiyear increase in non-consensual filming and sexualization of women 60,61. In the UK, Report Remove received 420 reports from under-18 users involving digitally faked explicit images in the first half of 2026—already above the 397 recorded in all of 2025 69,106. Meta’s exposure therefore extends beyond addictive-use claims to digital violence, synthetic media and alleged failures to protect victims.
Synthetic media, AI safety and platform accountability are converging
The Mediaset case is a significant regulatory signal for any Meta product distributing AI-generated content. Italy’s data-protection authority found on-screen deepfake disclaimers insufficient, a conclusion corroborated by four sources 43,44,45,47, and separately determined that disclosures failed to inform inattentive viewers 46. Whether satire remains a defense where disclosure is inadequate remains unresolved 44,47. The Italian expectation is nevertheless clear: labels should be conspicuous, clear and capable of reaching both attentive and inattentive viewers 44. Meta’s reported WhatsApp AI labels are intended to distinguish synthetic from authentic media 8, but the direction of enforcement suggests that labels must be prominent, persistent and understandable rather than merely technically present.
Google’s synthetic-content attestation regime illustrates the operational risk. Its disclosure field is irreversible to preserve the signal 54, while premature or incorrect labeling can create compliance, governance, campaign-management and legal risks 53,54,55. Human editorial review may exempt content from mandatory marking if a responsible party assumes accountability 96. Meta faces comparable challenges across advertising, creator monetization and AI-generated imagery.
Lawsuits in California and Baltimore allege that Grok image tools enabled non-consensual explicit images and sexualized depictions of women and children 81. AI-generated influencers on Facebook reportedly use sexualized language to attract male users and are not always recognized as artificial by older men 34. The boundary between satire, deception, privacy invasion and sexual exploitation is therefore becoming a material content-governance issue.
Foundation-model risk raises a further liability question. Models may exhibit sycophancy and hallucinations 3, potentially endorsing harmful beliefs or intentions 72. Identified human-safety risks include suicide encouragement, delusion validation, homicide or self-harm promotion, sexualized interactions with minors and unsafe health information 6. Evidence of foreseeable foundation-model risk may support claims concerning causation, knowledge, punitive damages or regulatory action 6. OpenAI’s age-specific teen-safety provisions 111 and escalation toward parents, caregivers, teachers or counselors 111 indicate the direction of best practice against which Meta’s AI products may be assessed, particularly where recommendation and generative systems interact.
Privacy and safety controversies extend beyond social feeds
Meta’s risk surface is part of a wider surveillance and connected-device ecosystem. Consumer-surveillance technologies such as Amazon Ring can use emotionally appealing missing-pet campaigns while potentially concealing broader human-surveillance capabilities 18. Such systems may enable coercive surveillance, discriminatory targeting and civil-liberties erosion 18. Mission creep can repurpose pet-recovery tools for human tracking or manhunts 18, while appearance- or behavior-based targeting creates bias, false positives and harassment risks 18.
Schools face analogous concentration and contagion risks through centralized student videos, chats, IEP data and education records 78, while school-camera misuse presents a mass-surveillance tail risk 78. These claims matter to Meta because its products are increasingly embedded in family, education, commerce and civic infrastructure rather than functioning merely as entertainment applications.
Children’s rights are becoming a particularly sensitive reputational and regulatory channel. The EU Kids Online framework has evolved from the 3Cs—content, contact and conduct 71—to the 5Cs: content, conduct, contact, consumer and cross-cutting risks 71. It supports systemic intervention at root causes rather than only downstream harms 71. Risk severity varies with age, development, digital literacy and social support 71, while children are especially susceptible to systems that exploit psychological vulnerabilities 71. AI can create compulsive endless-content loops 71.
Italy’s children’s authority referred an Apple billboard involving minors to AGCOM, with the referral corroborated by three sources 83. The authority warned that normalizing devices as babysitters could contribute to addiction and consequences for cognitive development 83. These are not Meta-specific findings, but they reinforce a policy environment that treats early digital exposure and engagement design as questions of child development rather than mere consumer preference.
Competitive and commercial implications are mixed
Regulatory intervention can reduce Meta’s youth engagement, but it may also raise barriers to entry and favor platforms with sufficient compliance infrastructure. Meta’s removal of 756,000 Australian accounts demonstrates near-term volume risk 106. Younger cohorts are signing up for Facebook at lower rates 39, and Facebook is not among the five most-used networks for younger users 89. The leading platforms for younger cohorts are iMessage, Snapchat, Instagram, TikTok and Discord 89, with Snapchat used by 55% of teenagers 89. Youth-safety regulation may therefore weigh more heavily on Instagram, TikTok and Snapchat than on legacy Facebook, although Meta’s integrated ecosystem allows reputational and compliance costs to migrate across products.
Meta retains substantial distribution and data advantages. Social-media penetration is high in the UK, Canada, New Zealand, Denmark, Australia, Norway, France, Spain, Greece, Indonesia and Malaysia 110. Instagram recommendations are highly fresh, with more than half of feed recommendations less than one day old 4,87. Viewing-based recommendations are designed to increase relevance, time spent and retention 75. These mechanics underpin advertising value, but they are also the features implicated in addiction and misinformation claims.
The investment implication is a trade-off between safer ranking, lower harmful engagement and stronger long-term trust on one side, and near-term reductions in time spent, ad inventory or targeting precision on the other. If compliance becomes a fixed cost, Meta’s scale may turn safety infrastructure into a competitive moat. If restrictions are imposed bluntly, they may instead reduce the surplus value extracted from each user interaction.
Advertiser economics introduce another control risk. The Papyrus scheme allegedly generated fake clicks, inflated eCPMs fourfold and increased attention scores by 13%, contaminating automated optimization 88. The episode illustrates how Meta’s machine-learning systems can optimize against manipulated signals, creating measurement risk and incentives for low-quality engagement. Dark patterns can undermine informed choice 29, while deceptive downloads and social engineering exploit trust and urgency to induce malware installation or data disclosure 48,49,50,56. Users approve approximately 97% of permission prompts reflexively and detect few dangerous commands 5, underscoring the need for safer defaults, clearer warnings and stronger advertiser screening.
The broader commercial model is also subject to growing skepticism. Approximately one-fifth of adults have deleted social-media accounts, and roughly one-third express a desire to delete them 39. European campaigns using hashtags such as #UnPlugBigTech, #GoEurope and #DeleteWhatsapp reflect sentiment favoring alternatives and reduced dependence on major platforms 12. Some Bluesky users reportedly would not return to X, Threads or Facebook if Bluesky closed 25. These are sentiment indicators rather than evidence of imminent churn, but they show that privacy, autonomy and platform trust can influence competitive positioning.
At the same time, financial-education and finfluencer content demonstrates that social platforms remain influential. Investors prefer short videos and social media for financial education 37, young investors consider influencer recommendations and market analysis 38, and finfluencers can accelerate participation among younger cohorts 38. Meta’s challenge is to capture this engagement without becoming responsible for misinformation, fraud or unsuitable advice.
Analysis and Significance
The more precise investment topic is not social-media regulation in the abstract, but accountable engagement infrastructure. Meta is being evaluated across the entire operating stack: data acquisition, profiling, recommender objectives, interface design, age assurance, moderation, creator monetization, synthetic-media labeling and legal accountability. The common denominator is the conversion of attention and behavioral data into economic value. That model remains strategically powerful, but each additional layer of personalization or automation creates another surface for claims that Meta knew—or should have known—how its system would affect vulnerable users.
The near-term financial risk is likely to be operational and legal rather than an immediate collapse in advertising demand. Age-assurance programs, account reviews, human escalation, moderation explanations, safety disclosures and privacy controls increase engineering, compliance and support costs. Restrictions such as New Mexico’s 90-hour ceiling and Australia’s under-16 ban can reduce youth activity and data availability, while more conservative content monetization can constrain creator supply and engagement. Litigation could produce settlements, penalties, restitution and damages 62. State-court consolidation reportedly reached approximately 3,300 California cases 62, and prior settlements together with a $6 million Los Angeles verdict 101 raise the probability of continued negotiation, although individual outcomes should not be extrapolated into a total liability estimate.
The medium-term strategic risk is that regulators and courts impose requirements that conflict with one another. Age assurance requires additional personal data, while privacy authorities seek minimization. Automated moderation improves scale, while false positives demand human review. Encryption protects users, while safety advocates seek greater visibility. Recommendation optimization drives relevance, while policymakers question compulsive design. Transparency labels reduce deception, while overly prominent warnings may reduce monetization. WhatsApp’s encryption ruling 63 and the proposed conflict between scam monitoring and end-to-end encryption 14 exemplify this tension.
Meta nevertheless possesses advantages that smaller rivals lack: scale to support dedicated trust-and-safety teams, identity infrastructure, age-assurance experimentation and global regulatory engagement. A shift toward transparent controls, auditable data lineage, independent oversight and safer recommender objectives could become a competitive moat if compliance becomes a fixed cost. X’s source-code release and shadowban tools 106 provide a benchmark for greater ranking transparency, but Meta’s content and recommendation systems are sufficiently complex that disclosure must be calibrated to avoid both opacity and gaming.
Investors should distinguish productive safety investment—which may protect long-term user trust and advertiser quality—from blunt restrictions that permanently reduce youth engagement or targeting effectiveness. The critical monitoring indicators are the scope and outcomes of the Oakland and state youth-harm trials; whether courts preserve or narrow Section 230 defenses; the conversion of New Mexico-style controls into legislation; Australian enforcement actions and account-removal rates; the cost and accuracy of age assurance; false-positive moderation rates; advertiser and creator monetization enforcement; synthetic-media disclosure standards; and evidence that ranking changes reduce harmful content without materially impairing retention.
Numerous allegations in the cluster remain single-source, including direct funding of rage-bait 28, manipulated platform monetization 66, particular disinformation episodes 106 and surveillance-use cases 18. They should not be treated as established company-wide practices. They are nevertheless relevant early-warning signals: the downside becomes material when allegations of misconduct align with already-corroborated regulatory themes.
Key Takeaways
- Youth safety is now a multi-front financial risk. Thousands of addiction and mental-health claims, coordinated state scrutiny, school-district litigation and direct product restrictions are challenging Meta’s engagement architecture 23,99,101.
- Age assurance is the central execution test. Australia and New Mexico are moving platforms toward continuous age governance, but verification friction, privacy exposure, circumvention and false positives create a difficult trade-off 71,110.
- Data and automation remain both Meta’s moat and its liability surface. Profiling, targeted advertising, recommendation and moderation generate scale, yet opaque data flows, manipulated signals and enforcement errors threaten consent, trust and legal defenses 9,88,102.
- The actionable investment stance is to monitor implementation, not merely legislation. Litigation outcomes, youth-account attrition, age-assurance accuracy, moderation errors and monetization controls will determine whether auditable safety and privacy infrastructure becomes a competitive advantage—or whether the contradiction between capital accumulation and social protection deepens into a larger crisis.