we_are_coded.by CODE · The world, decoded
БГ
Who's who

How America regulates AI: the FTC waits, the states lag

The BasicsUpdated on 16 August 2026we are coded

In America, artificial intelligence ships first and explains later - if it even gets caught.

Checked on16 August 2026
In short: the US has no single law for artificial intelligence. The Federal Trade Commission (FTC) - the agency that protects consumers from fraud - steps in only after something's gone wrong, not before. Individual states write their own rules, each at its own pace. Colorado went first, but its own law has been delayed three times and still hasn't taken effect.

The FTC hasn't written a new law for AI - it's just stretching an old one. The section of the law that governs it bans "unfair or deceptive practices", and as of July 2026 the commission argues that systems secretly tuned to skew their answers fall right under this. You don't have to lie outright. It's enough to tweak what the algorithm shows, without telling anyone.

In practice this means: a chatbot that doesn't admit it's a bot is in violation. An AI trained to stay silent on certain topics or present them in a distorted way, without the user knowing, is also in violation. The FTC looks at the result on screen, not the code behind it.

Congress hasn't passed a federal AI law, and nothing suggests it will anytime soon. So the vacuum gets filled from below: state by state, with different rules and different deadlines, instead of one common framework from above.

Colorado tried to be the first state with a real AI law, passed back in 2024. But every time its effective date got close, someone pushed it back. The latest version moves it to 2027, and along the way pulled most of its teeth: the mandatory risk assessments disappeared, and so did the general obligation for companies to protect consumers from algorithmic discrimination. What's left is mostly a sign saying "disclose what your AI-system does", without much behind it.

California and Illinois have their own, narrower laws - on hiring discrimination or on disclosing that you're talking to a bot. The result is a patchwork of laws, not a finished picture. A company in San Francisco follows one set of rules, its neighbor in Texas - almost none.

Sue after the fact, not before it.

Here's where it limps

The American model isn't laziness. It's a bet. The bet is that the market and the courts will correct faster than officials writing rules in advance for a technology that changes every month. The problem is, while you wait for the case that proves the harm, the harm has already happened to a specific person - and they don't get a law to protect them in advance, only a chance to sue afterward, if they even find out what happened to them.

So I wouldn't call this "lighter regulation". I'd call it regulation looking backward in time - you look in the rearview mirror and call it radar.

The visual is generated code art. No third-party images.
Official primary sources
→FTC: Artificial Intelligence - what the commission does (official)→Colorado General Assembly: SB24-205 - Consumer Protections for AI