48 US States, 478 AI Laws. One Rule Your Business Can't Ignore

In Episode 1 of this series we covered the rule that a chatbot must tell people it is AI, and in Episode 2 the rule that AI-generated content must be labelled. Both came from the EU AI Act, and a fair question is whether they matter to a business that does not operate in Europe. This update is the answer: the same principles are spreading well beyond the EU, most visibly across individual US states, and the direction of travel is unmistakable.

A wave of state laws, not one headline

The real story is not a single dramatic law but the sheer volume of them. Trackers now count hundreds of AI transparency bills moving through nearly every US state, covering watermarking and provenance, chatbot disclosure, synthetic media labelling, and training-data transparency. Dozens are taking effect during 2026. For a business, this is arguably harder to navigate than one big federal law would be, because it is a patchwork: different states, different rules, different dates. But underneath the variation, the core ideas are strikingly consistent, and they are the same ones this series has been teaching.

Chatbot disclosure is going global

The rule from Episode 1, that an AI must disclose it is an AI, is exactly the kind that is spreading. Several US states are advancing bills requiring businesses to tell people when they are communicating with artificial intelligence rather than a human, the same core duty as the EU's. There is also movement at the federal level: proposed legislation would require consumer AI agents to disclose their non-human status nationally. Nothing about this is uniquely European any more. If your chatbot serves customers in multiple markets, the safe assumption is that a disclosure duty either already applies or soon will, wherever they are.

Content labelling and watermarking too

Episode 2's theme, labelling AI-generated content, is following the same path. California's watermarking law, described by trackers as among the most comprehensive, requires generative AI providers to embed hidden disclosures in AI-generated images, the machine-readable marking side we discussed in our watermarking update. Other states are targeting synthetic media in advertising and political content. The specifics vary, but the principle is the one this series keeps returning to: people are entitled to know when what they are seeing or reading was made by AI.

The pattern worth understanding: disclosure survives, bans do not

Here is the most useful insight for anyone trying to read where this is heading. Not every AI law survives contact with the courts, and there is a clear pattern in which ones do. Broad bans on AI content have repeatedly run into trouble: a sweeping California law was permanently blocked in 2025 after a federal judge found it likely unconstitutional on free-speech grounds, and a Hawaii law met a similar fate. But disclosure-based laws, the kind that require transparency rather than prohibition, have generally held up. The lesson is that the durable approach, the one likely to define the landscape, is exactly the transparency-and-labelling model at the heart of the EU rules and this series. Betting on transparency is betting on the version of these rules most likely to last.

What this means for you

Three practical points. First, do not dismiss AI transparency rules as a European issue, the same duties are arriving across the US and elsewhere, and if you serve customers in several places, some version very likely reaches you. Second, because it is a patchwork of overlapping rules rather than one clean law, the simplest safe strategy is to meet the strictest common standard: disclose your AI, label your AI content, and you are broadly covered across jurisdictions at once. Third, this is not slowing down, so the businesses that build transparency in now will find themselves ready as more rules land, rather than scrambling each time. Next week we return to the numbered series with how to classify your AI system into the right risk tier.