where are we at the moment with AI governance

Early internet governance mistakes—fragmented regulation, siloed stakeholder input, and reactive policymaking—offer clear lessons for AI lawmaking. - Inclusive, multi-stakeholder design: Internet rules were often shaped by narrow technical or commercial interests, producing blind spots (e.g., privacy and content harms). AI laws should involve governments, technologists, civil society, affected communities, and independent experts from the start to surface diverse risks and values. (See: Berners-Lee on governance; multi-stakeholder models in ICANN history.) - Principle-driven but operationalized rules: Broad principles (free speech, innovation) proved insufficient without operational definitions and enforcement mechanisms. AI regimes need clear standards, measurable compliance requirements, and practical audits, not only high-level ideals. (Compare: GDPR’s rights + enforcement vs. early net norms.) - Anticipatory and flexible regulation: The internet’s reactive patchwork allowed harms to scale before remedies arrived. AI laws should be adaptive, include sunset/review clauses, and enable rapid updates as capabilities and harms evolve. Regulatory sandboxes can allow experimentation while limiting systemic risk. - Interoperability and cross-border coordination: Fragmented national rules created compliance burdens and safety gaps. International coordination on baseline norms, export controls, and data standards reduces regulatory arbitrage and improves safety clustering. (See: Budapest Convention, GDPR influence.) - Accountability, transparency, and incentives: Without clear accountability mechanisms, platforms optimized growth over safety. AI law should align incentives—mandate transparency, independent audits, incident reporting, and proportionate penalties—to make compliance feasible and meaningful. - Equity and access considerations: Early internet policy sometimes prioritized infrastructure and markets over equitable access and protections for marginalized users. AI governance must foreground distributive effects and protect vulnerable populations from bias and surveillance. Taken together, these lessons point to laws that are inclusive, operational, flexible, internationally coordinated, and enforcement-ready—so we don’t repeat the internet’s governance shortfalls when regulating AI.