A new survey finds that half of American adults used an AI service in the past week. ChatGPT is the most common entry point, but the category has expanded well beyond a single app: AI assistants, AI search, AI shopping agents, AI writing tools. The behavior has quietly normalized in the span of roughly two years, spreading from early adopters to a population that mostly did not know what a large language model was eighteen months ago.
Congress, meanwhile, has produced essentially nothing in the way of federal AI legislation. What has emerged from Washington is a patchwork of executive orders, agency guidance documents, voluntary commitments from tech companies, and interagency task forces that report to other interagency task forces. This is not regulation. It is the performance of regulation.
The gap between the pace of adoption and the pace of governance is not new in American technology policy — we have been here before with social media, with algorithmic recommendation, with data privacy. The difference this time is the speed. Social media took roughly a decade to go from novelty to infrastructure. AI is doing it in two or three years, and the downstream effects — on labor markets, on information, on the meaning of expertise — are correspondingly faster.
The argument against moving quickly is always the same: we do not yet understand the technology well enough to regulate it without stifling innovation. This argument has the advantage of being partially true and the disadvantage of being infinitely extensible. There will always be something we do not yet understand. At some point the absence of understanding becomes a choice, not a condition.
Half of America is already living with AI. Washington should probably catch up.