How to regulate AI

Artificial intelligence is not just transforming industries; it's reshaping the very fabric of our society, from how we work and learn to how we connect and care for ourselves. As this powerful technology accelerates, a critical question emerges: how do we govern AI responsibly, ensuring its benefits uplift humanity while mitigating its profound risks? The choices we make today in regulation will define our collective future with AI.

Navigating AI's Financial Frontier

The ubiquity of AI in business and finance introduces unprecedented risks. Algorithms, designed to optimize profits, can tacitly learn to collude, inflating prices without direct human instruction. This raises complex questions about responsibility in antitrust cases, an area where current legal frameworks fall short.

Combatting AI-Powered Scams and Autonomous Transactions

AI's persuasive capabilities already surpass skilled human negotiators. This power, when applied to vulnerable populations, can transform traditional scams into highly personalized, sophisticated schemes. Imagine deep-fake audio and video deployed en masse, deceiving even the most cautious individuals. More alarming still is the prospect of AI agents with direct access to immutable financial systems, such as cryptocurrency networks. An AI instructed to “grow its portfolio” could deploy fraudulent smart contracts or transactions that no authority can halt or reverse. To counter these emerging threats, we need enhanced crypto monitoring, mandatory “kill switches” for AI agents, and human-in-the-loop requirements for critical models. Addressing these challenges demands immediate, collaborative action between innovators and governments.

Choosing Our AI Future: The Pluralism Paradigm

Globally, three main paradigms for governing AI are emerging. The “accelerationist” model prioritizes speed and technological development, potentially replacing human labor and centralizing success among high-IQ individuals. The “effective altruism” paradigm also embraces rapid development but plans to mitigate its societal impact with universal basic income. However, a third path, “pluralism,” offers a more human-centric vision.

Empowering Human and Machine Intelligence Together

The pluralism paradigm focuses on AI complementing and extending the diverse forms of human intelligence, rather than overmatching or replacing them. This approach aims to activate creativity, innovation, and cultural richness, while fully integrating the broader population into the productive economy. Examples like Pennsylvania's commitment to technology that empowers rather than replaces, or Utah's Digital Choice Act placing data ownership back with users, illustrate this path. If we are to uphold democratic values and individual freedom, pursuing the pluralism paradigm is not merely an option, but a necessity.

Designing Safe AI for Mental Well-being

As more people, particularly teens, turn to AI for mental health advice and emotional support, regulation must balance reducing harm with promoting access to evidence-based resources. People will continue to ask chatbots sensitive questions; the goal is to make these interactions safer and more useful.

Essential Guardrails for AI Mental Health Support

Regulatory priorities should include requiring standardized, clinician-anchored benchmarks for suicide-related prompts, ensuring public reporting. These benchmarks must include multi-turn dialogues to test nuanced scenarios where chatbots could inadvertently cross a “red line.” Strengthening crisis routing with up-to-date, geolocated resources and “support-plus-safety” templates is crucial. Privacy must be enforced, prohibiting advertising and profiling around mental health interactions, minimizing data retention, and requiring a “transient memory” mode for sensitive queries. Any model marketed for mental health support must meet a duty-of-care standard through pre-deployment evaluation, post-deployment monitoring, independent audits, and alignment with risk-management frameworks. Finally, independent research funding is vital to ensure safety tests keep pace with rapid model updates. Setting a high floor of safety and transparency now will enable trusted, more comprehensive mental health functions in the future.

The Imperative of Global AI Collaboration

Current AI policy discussions are often framed by geopolitical competition, viewing technological advancement as a zero-sum game. This perspective can obstruct the coordination needed for global AI safety frameworks and dialogues. The history of AI development, with its international leading teams, underscores the value of collaboration. While maintaining U.S. innovation dominance is important, we must recognize that AI products developed in one hub may not suit all global applications. Fostering local collaborations and entrepreneurship ensures AI technology is relevant to local contexts and reaches a global audience. Paradoxically, ceding more control to diverse global development could, in fact, consolidate technology and market power for U.S. AI innovators by making their solutions more universally adaptable and trusted.

Balancing Innovation with Accountability

The rapid adoption of AI is unfolding amid a heightened awareness of the societal implications left unchecked by previous technology waves, such as internet platforms and social media. There is a strong push to repeat that “light on safeguards, rich in incentives” playbook. However, innovation and accountability are not trade-offs; they are a dual imperative. Dismissing guardrails as barriers to innovation leaves critical questions unanswered: Who ensures fairness in algorithmic decision-making? How do we protect workers displaced by automation? What happens when infrastructure investment prioritizes computing power over community impact?

Building Trust through Transparency and Oversight

While supporting infrastructure and workforce development is essential, we must also incentivize standards-based independent red-teaming, support a robust market for compliance and audits, and build governmental capacity to evaluate AI systems effectively. If the world is to trust American-made AI, we must ensure it earns that trust through rigorous, transparent accountability, both at home and abroad.

Smarter Regulation for Healthcare AI

Current clinical AI regulation often misses the mark, being mismatched to the real-world problems clinicians face. Efforts to fit AI into existing device pathways can narrow its application, reduce its perceived risk, but ultimately suppress its impact and adoption. This approach fails to address the true bottlenecks in U.S. care, such as efficiency under rising patient volumes and workforce shortages. Foundation models can draft reports, summarize charts, and orchestrate routine workflows, yet a widely used pathway for these continuously learning clinical copilots is still absent.

Focusing on Post-Deployment Monitoring and Real-World Evidence

While some pre-market requirements might eventually lighten, more responsibility will inevitably shift to developers and deploying providers. This shift is only feasible if providers have practical tools for local validation and continuous monitoring after deployment, as most are already overwhelmed. Instead of solely relying on pre-market assessments, regulation should welcome approaches like “regulatory sandboxes” that allow for rapid, supervised testing in real settings, generating evidence for agencies and payers alike. A crucial step is to require local validation before implementation, continuous post-market monitoring — through registries like the American College of Radiology’s Assess-AI — and routine reporting back to regulators. This allows for observation of effectiveness and safety in practice, moving beyond theoretical generalizability. Healthcare AI needs policies that expand trusted, affordable compute, adopt robust monitoring, enable sector testbeds at scale, and reward demonstrated efficiency to protect patients without impeding progress.

The conversation around AI regulation is no longer theoretical; it's an urgent, practical necessity. From protecting financial systems and mental well-being to shaping our societal values and global collaborations, the stakes are incredibly high. By embracing a proactive, thoughtful, and human-centric approach to governance — one that champions transparency, accountability, and collaboration — we can harness AI's immense potential to build a more equitable, innovative, and resilient future for all. The opportunity is now; let's collectively define the path forward.