Instead of trying to anticipate every potential harm, AI regulation should mandate open, internationally consistent audit trails, similar to financial transaction logs. This shifts the focus from pre-approval to post-hoc accountability, allowing regulators and the public to address harms as they emerge.

The traditional government model of setting a regulation and waiting years to assess it is obsolete for AI. A new approach is needed: a dynamic board of government, industry, and academic leaders collaborating to make and update rules in real-time.

The White House's proposed legislative framework explicitly recommends against creating a new, overarching federal body to regulate AI. Instead, it advocates for empowering existing agencies with subject-matter expertise (e.g., in finance or healthcare) to develop and enforce AI rules within their own domains, suggesting a decentralized approach to governance.

India is taking a measured, "no rush" approach to AI governance. The strategy is to first leverage and adapt existing legal frameworks—like the IT Act for deepfakes and data protection laws for privacy—rather than creating new, potentially innovation-stifling AI-specific legislation.

Learning from its failed comprehensive AI bill, the Canadian government is adopting a step-by-step strategy. It's first addressing overdue updates to privacy laws and online harms like deepfakes. This pragmatic approach builds a solid legal foundation before attempting another ambitious, all-encompassing AI bill, making progress more politically viable.

Creating a new regulatory framework for AI is premature because nearly all feared harms—from malpractice to non-consensual imagery—are already illegal under existing laws. The initial focus should be on applying these established laws to AI-assisted actions, not inventing a new regime from scratch.

Instead of only using AI to help people comply with complex regulations, its real power lies in helping policymakers simplify them. AI can analyze thousands of pages of rules to identify what is vestigial, conflicting, or redundant, enabling the simplification required for scalable government services.

Overly-specific regulation focused on AI tools (e.g., model size) risks accidentally stifling valuable, unforeseen use cases. A better policy focuses on outcomes. For example, prosecute fraud committed with an LLM, but don't regulate the LLM itself, thereby protecting innovation while punishing misuse.

A16z advocates for a "gap analysis" approach to AI regulation. Instead of assuming a legal vacuum exists, lawmakers should first examine how existing, technology-neutral laws—like consumer protection or civil rights statutes—already apply to AI harms. New legislation should only target clearly identified gaps.