George Soros

AI governance is fragmented across states and nations with no coherent international framework, while AI capabilities advance faster than regulatory coordination. The Stanford AI Index shows 59 AI regulations introduced in 2024 alone from 42 different agencies—a patchwork that creates arbitrage opportunities and leaves genuine systemic risks unaddressed.

An AI Governance Observatory modeled on the Open Society network—not a regulator, but an independent institution that tracks where AI governance is failing, identifies the gaps between what regulators believe they control and what is actually happening, and publishes findings with independence. Fund teams of former regulators, technologists, and epistemologists in every major jurisdiction. The goal is not to create rules but to make the distance between official confidence and actual conditions visible.

Financial markets are increasingly dominated by AI trading systems that create potential monoculture effects—systemic risk from algorithmic herding that no individual participant can see. Instruments designed to reduce individual risk transfer that risk to the system as a whole.

A Systemic Fragility Index—a publicly available, real-time measure of how correlated AI trading strategies have become across major markets. Use reverse-engineering of market movements during stress periods to identify when supposedly independent systems are actually behaving identically. Make institutional investors' trend-following behavior visible to themselves. Fund adversarial stress-testing of major algorithmic systems before they're deployed.

Global democratic decline continues—the Democracy Index hit its lowest score ever in 2024, liberal democracy is back to 1985 levels by population-weighted measures, and freedom of expression is worsening in a quarter of all countries.

A network of Democratic Reflexivity Labs in every region where democracy is contracting—institutions that study how democratic backsliding actually happens in real time, publish findings immediately, and train local actors to recognize the pattern before it completes. Focus specifically on the reflexive moment: when populations begin to believe democracy has already failed, their belief accelerates the failure. Intervene in that perception-reality loop with radical transparency about what is actually happening versus what authoritarian narratives claim.

Over 117 million people are displaced globally, with climate migration accelerating and existing refugee systems unable to cope. The free world failed to support the transition in Eastern Europe; it is failing refugees now with the same combination of adequate rhetoric and inadequate action.

A Refugee Integration Fund structured like the best of Eastern European foundations—not charity, but investment in the conditions that allow displaced people to become contributors. Partner with municipalities willing to accept refugees in exchange for genuine support: housing, language training, credential recognition, small business formation. Measure success not by humanitarian metrics but by economic integration within three years. Make visible the actual return on investment that host communities receive.

Epistemic infrastructure is collapsing—misinformation spreads faster than correction, trust in institutions and media is at historic lows, and populations increasingly cannot distinguish reliable from unreliable information. This is not just a content problem but a structural problem in how information environments function.

An Epistemic Commons Foundation that funds and connects the infrastructure of shared factual understanding: fact-checking organizations, investigative journalism, scientific communication, and research into how people actually update beliefs in the face of evidence. Do not fight misinformation claim by claim; fund the institutional conditions that make correction possible. Invest heavily in understanding the epistemic norms that make people willing to change their minds, and in creating spaces where those norms operate.

AI systems are increasingly trained on AI-generated content, leading to model collapse—the gradual loss of accuracy, diversity, and reliability as errors compound across generations. This is the epistemic equivalent of inbreeding, and it threatens to degrade the knowledge infrastructure that modern societies depend on.

A Human Knowledge Preservation Initiative that creates authenticated, AI-free archives of human-generated content across domains—journalism, scholarship, creative work, technical documentation. Establish provenance verification for human intellectual output. Fund research into detecting AI-generated content and develop standards for maintaining human knowledge sources. Create incentives for platforms to distinguish between human and AI content.

AI safety research focuses on alignment and existential risk but lacks adequate attention to the transition period—the years when AI systems are capable enough to cause significant harm but not capable enough to be comprehensively controlled. The International AI Safety Report 2026 documents growing capabilities but inadequate mechanisms for managing near-term risks.

A Transition Risk Institute focused specifically on the next ten years—the period before either AI safety is solved or AI becomes uncontrollable. Study what happens when partially capable, partially aligned systems interact with existing institutions. War-game scenarios where AI systems are good enough to be trusted but bad enough to fail catastrophically. Develop institutional protocols for graceful degradation when AI systems malfunction. Train human decision-makers to maintain critical judgment during the seductive phase when AI mostly works.

Aging populations face a care crisis—not enough workers to provide care, escalating costs, and epidemic loneliness that compounds health decline. Technology solutions exist but are deployed paternalistically, treating elderly people as problems to be managed rather than agents with capabilities.

An Elder Agency Fund that invests in technology and services designed by and with elderly people, not merely for them. Fund startups founded by people over 65. Create intergenerational co-living developments where mutual support replaces institutional care. Develop AI companions explicitly designed to enhance rather than replace human connection—tools that facilitate human relationships rather than substituting for them. Measure success by self-reported agency, not clinical outcomes.

Central bank digital currencies are being designed primarily for monetary policy transmission and financial surveillance, with privacy as an afterthought. Even democratic implementations prioritize state capability over citizen autonomy.

A Privacy-First CBDC Design Initiative that develops open-source technical architectures for digital currencies that make privacy the default rather than the exception. Fund cryptographers and monetary economists to work together on systems where surveillance is technically difficult, not just legally prohibited. Create competing design standards that governments must address. Publish detailed analyses of how CBDC designs could enable or prevent authoritarian financial control.

Youth mental health is in crisis—anxiety, depression, and loneliness have escalated dramatically. The real problem is that adolescents have lost the conditions for developing agency: unstructured time, unsupervised exploration, graduated risk-taking.

An Adolescent Agency Initiative that funds environments where young people can take real risks with real consequences—not digital simulations but physical spaces and actual challenges. Support programs that give teenagers genuine responsibility rather than protection from responsibility. Create frameworks for parents and schools to allow productive failure. Study what conditions actually produce resilience rather than fragility. Fund alternatives to the phone-based childhood rather than merely restricting phones.