Europe’s legal frameworks lag behind AI‑amplified online harms against children, leaving regulatory gaps that enable cross‑border hosting of abuse and weak private incentives to prevent harm; robust, technology‑aware, rights‑preserving, and internationally coordinated rules — alongside enforcement and clear platform obligations — are needed to align market incentives and protect young users.

Main Finding

Current European legal and regulatory frameworks provide important building blocks for protecting minors online but are fragmented, unevenly enforced, and not fully fit for purpose in the age of generative AI. New risks—especially AI-generated child sexual abuse material (AI‑CSAM), deepfakes, and scale effects from algorithmic amplification—require targeted, harmonized legal responses (including classifying certain generative models as high‑risk), clearer platform duties, improved age‑assurance and risk‑assessment frameworks, and coordinated cross‑border enforcement. Without such adaptation, harms to children will likely increase while compliance and mitigation costs for platforms and states will grow.

Key Points

• Digital immersion of youth is near‑universal in Europe (e.g., 97% daily internet use among young people in the EU, strong social media uptake), producing both educational/creative opportunities and pronounced exposure to harms.
• Range of online harms analyzed: CSAM (including AI‑CSAM), cybergrooming, sexting/sextortion, cyberbullying, hate speech, exposure to inappropriate content, addiction, extremism/radicalization, algorithmic harms, risky behavior, and data breaches.
• AI introduces novel modalities of harm (synthetic CSAM, deepfakes, automated grooming tools, hyper‑targeted exploitation) and complicates detection, attribution, and remediation.
• Comparative legal review covers international instruments (UN, Council of Europe, OECD), EU strategies (BIK+, GDPR, EU guidance on age assurance), and national measures across many European states (e.g., Germany NetzDG, France kidfluencer rules, UK Online Safety Act, proposed Swiss frameworks).
• Institutional responses surveyed: watchdogs, UNICEF guidance (including AI for children), national data protection authorities (AEPD, CNIL), Safer Internet Centres, and NGOs.
• Platform action is documented (policies and feature changes across Instagram, TikTok, YouTube, Google), but platform measures are often insufficient, inconsistent, or non‑transparent.
• Identified regulatory gaps: fragmentation across jurisdictions, tensions between privacy and surveillance‑style remedies (e.g., chat control proposals), limited effectiveness of current enforcement, lack of standardization for age assurance, and insufficient rules for generative AI.
• Policy recommendations recurring in the book: mandatory risk assessments, classifying generative AI with child‑harm potential as high‑risk, mandatory platform training and content moderation standards, broadened definitions of inappropriate content, better cooperation with law enforcement, promotion of media literacy, and proportional age assurance mechanisms.

Data & Methods

• Methodology: doctrinal legal analysis and comparative law approach. The authors examine statutes, regulations, guidance documents, national legislative proposals, platform terms of service and technical measures, institutional guidance (UNICEF, OECD, Council of Europe), and enforcement practices.
• Empirical context and secondary data: integration of public statistics and social‑science studies to show scale and shape of youth online usage and harms (Eurostat usage rates; surveys such as UNICEF Switzerland & Liechtenstein; MIKE and JAMES studies on youth online experiences).
• Case‑based policy review: analysis of specific national laws, platform practices (Instagram, TikTok, YouTube, Google), and proposals (e.g., chat control), together with evaluation of regulatory effectiveness and institutional capacities.
• No primary quantitative modeling of economic effects in the book; conclusions about costs and impacts are drawn from legal texts, policy documents, platform disclosures, and academic/NGO studies.

Implications for AI Economics

• Compliance and operational costs
  • Classifying certain generative AI models as high‑risk will raise compliance costs for model developers and platform operators (risk assessments, documentation, audits, technical mitigation, possible licensing/certification).
  • Mandatory age‑assurance and safety-by-design obligations will require investment in verification technologies, identity/privacy tradeoffs, and integration costs—raising barriers to entry for smaller firms and increasing fixed costs for incumbents.
• Moderation economics and labor
  • The scale and sophistication of AI‑enabled harms increase demand for both automated detection tools and specialized human moderators. This will shift cost structures: higher up‑front R&D and tooling costs plus ongoing human review costs to manage false positives/negatives and nuanced cases involving minors.
  • Overreliance on automated systems risks overblocking (reducing content supply and engagement) and underblocking (allowing harms), each with economic externalities for platforms and users.
• Market structure and competition
  • Stricter, harmonized regulation may benefit large platforms that can shoulder compliance costs while disadvantaging startups and smaller firms—potentially increasing concentration in the platform market.
  • Conversely, compliance requirements create markets for safety and verification services (age‑assurance vendors, automated CSAM detection providers, auditing/certification firms), producing new entrants and competitive niches.
• Advertising, targeting, and business models
  • Privacy and child‑safety rules (GDPR interplay, restrictions on profiling minors) constrain high‑precision ad targeting, affecting ad revenues derived from youth demographics. Platforms may need to adapt monetization strategies (contextual ads, subscription models, or age‑segmented offerings).
  • Limits on data use for minors could reduce the marginal value of young audiences to advertisers, influencing content strategies and platform investment in youth‑oriented features.
• Innovation incentives and risk allocation
  • Liability regimes and mandatory safeguards influence incentives for deploying generative AI capabilities: stricter liability or high‑risk classification discourages risky model releases but may also slow beneficial innovation (e.g., age‑appropriate educational AI).
  • Clear, predictable regulation (harmonized across jurisdictions) lowers regulatory uncertainty, which can support responsible investment, whereas fragmented rules raise compliance complexity and cross‑border frictions.
• Enforcement externalities and cross‑border coordination
  • Effective cross‑border enforcement is essential because harms and platforms are transnational. Fragmentation raises enforcement costs and risks regulatory arbitrage; harmonization reduces duplication and provides economies of scale for compliance.
• Measurement and social welfare
  • Economic welfare calculations must account for non‑market harms (psychological, developmental) from exposure to AI‑enabled harms. Investing in prevention (education, safer default settings, platform design) may have high social returns but require up‑front public and private spending.
• Opportunities for public‑private partnerships
  • Governments can subsidize detection and remediation R&D, fund media literacy programs, and support certification schemes to internalize externalities and reduce private costs of ensuring child safety.
• Policy design tradeoffs
  • Regulators must balance false positive costs (censorship, lost engagement and creator income) against false negative costs (harm to children, reputational and legal costs). Optimal policy mixes will combine technical standards, human review, liability rules, and demand‑side measures (education).
• Research and market data needs
  • Better microdata on moderation performance, false positive/negative rates, economic impact on revenues/usage, and compliance costs are necessary for efficient policy design. The book highlights empirical gaps that economists can help fill.

Concluding note: The book offers a comprehensive legal map of current protections and gaps. For AI economics, its principal value is identifying where regulatory changes will reallocate costs and incentives across firms, technologies, and public actors—suggesting priority areas for economic analysis (costs of compliance, moderation labor markets, platform competition effects, and welfare tradeoffs from safety interventions).