What If We Regulated AI Like Cars?

Summary

This opinion essay argues that automobile regulation could provide a practical template for managing powerful AI systems that write and modify code. It begins from the concern that coding systems and agent swarms could damage digital infrastructure, while overly restrictive rules could weaken the position of the jurisdiction that adopts them. The proposed framework would require users of sufficiently powerful coding AI to pass a basic licensing test and register each system, after which operation would be permitted under rules designed to protect society. Operators would generally be responsible for damage caused by their systems, while manufacturer liability could apply when a central defect resembles a recalled vehicle or faulty part. Insurance could help compensate victims without making every individual error financially ruinous. Before public use, AI models could face safety and environmental-impact standards analogous to vehicle safety and emissions tests. The essay also suggests age limits, license revocation, registration requirements for self-trained systems, and penalties for operating unregistered systems. Registration and licensing fees could support public infrastructure or assistance programs. The author argues that automobile regulation offers a century of practical testing, legal mechanisms, flexibility across jurisdictions, and room for commercial research and development rather than a wholly untested regulatory scheme. Regulation could also create jobs in compliance, insurance and human oversight, with people remaining available to resolve cases automated systems cannot handle. The proposal is presented as an analogy and starting point, not as a detailed legislative plan, and the essay closes by noting that existing vehicle regulation is imperfect while still allowing useful and innovative systems to coexist with public safety.