An Open Letter to Government: Leave AI Alone
To the Members of Congress, Regulatory Agencies, and Policymakers,
We write to urge restraint. As you consider new rules for artificial intelligence, we ask you to resist the instinct to regulate first and understand later. The history of transformative technologies — the printing press, electricity, the internet — teaches the same lesson every time: premature regulation calcifies incumbents, chills experimentation, and rarely protects the people it claims to serve.
Innovation needs room to fail. AI is still in its formative years. Model architectures, safety techniques, and deployment patterns are changing month to month. Rules written today, based on today’s models, will be obsolete — or actively counterproductive — by the time they take effect. Static law cannot keep pace with a moving target, and rigid compliance regimes built around current fears will lock in yesterday’s assumptions.
Heavy compliance burdens favor giants, not the public. Every licensing requirement, every mandatory audit, every certification regime raises the fixed cost of building AI. Large incumbents can absorb that cost easily; startups, open-source developers, and independent researchers cannot. The practical effect of aggressive regulation is not safer AI — it’s fewer competitors, less scrutiny of the biggest players, and a handful of companies deciding the technology’s future on your constituents’ behalf.
America’s edge is fragile. Global development of AI is not waiting for any single country’s approval. Capital, talent, and compute are mobile. A jurisdiction that regulates ahead of its peers doesn’t make AI safer worldwide — it just exports its own AI industry elsewhere, while ceding influence over the standards that end up governing the technology globally.
Existing law already covers most harms. Fraud is illegal. Discrimination is illegal. Defective products create liability. Consumer protection statutes already apply regardless of whether a computer program was involved. Before writing new AI-specific statutes, regulators should demonstrate that current law is actually insufficient — not simply that AI is unfamiliar.
Ask for transparency, not permission. Where government has a legitimate role, it is in requiring disclosure — what a model was trained on, how it’s evaluated, where it’s deployed in high-stakes settings like hiring, lending, and healthcare. Sunlight lets markets, journalists, and courts do their work. Pre-approval regimes and licensing schemes do not.
We are not asking for a lawless frontier. We are asking you to legislate narrowly, based on demonstrated harms rather than speculative ones, and to leave the space for competition and iteration that has made every prior wave of American technology leadership possible.
Regulate the harm. Not the technology.
Respectfully, Concerned Builders and Investors in AI