Brad Smith is right, and the fact that his comparison feels almost absurd is exactly why it matters. Milk and washing machines are subject to more oversight than the systems now shaping hiring, lending, and medical advice. That is not a metaphor. It is a description of a regulatory vacuum that has persisted long enough to become a joke about our own priorities.
We should be careful not to mistake the mundane comparison for a modest ask. Smith is not suggesting we treat AI like a perishable dairy product or a spin cycle. He is pointing out that we already decided, as a society, that certain technologies deserve guardrails because they can hurt people in predictable ways. A defective washing machine can flood a home. Contaminated milk can poison a family. An AI system making consequential decisions about a person's life can do damage that is slower to surface but no less real. We built oversight for the first two because we understood the risk. We have not built it for the third because the risk is still abstract to most of us.
What Smith is really calling for is a shift in how we frame the debate. Instead of asking whether AI is dangerous enough to regulate, we should ask whether it is consequential enough to leave unregulated. The answer is obvious once you stop treating the technology as a novelty. This is not about stifling innovation. It is about applying the same logic we use for appliances to tools that already influence everything from how AI-powered balloons earn military contracts to how developers rebuild familiar software from scratch. Those examples show how quickly AI moves from experimental to operational. The military is betting on it for weather prediction. Developers are using it to recreate products we rely on daily. The technology is already embedded in infrastructure, and our rules have not caught up.
The practical consequence for readers is not a distant policy debate. It is about accountability. When a washing machine fails, you know who made it and who is responsible. When an AI system makes a bad decision, the chain of responsibility is often murky, spread across model developers, deployment teams, and the companies that bought the tool. Regulation would force that chain to become visible. It would require that someone answer for outcomes, not just outputs.
We should watch what Microsoft does next, not just what Smith says. A company that profits from AI infrastructure asking for regulation is a signal that the industry knows the current state is unsustainable. The open question is whether the rules will be shaped by the people who build the systems or by the people who live with their consequences. That distinction matters more than any timeline for legislation. The milk and washing machines did not write their own rules. Neither should AI.