Yesterday, I discovered the Federal Register. It wasn't hiding or anything. My high school simply neglected to tell me about the constitutional process in favor of making me remember the preamble. Anyway, if you're just now learning about the Federal Register... you're welcome. Now, what does that have to do with AI? Usually, not much. But it's the place where the federal government does a bunch of stuff I don't understand, and, crucially, where people like you and me get to comment on what the different agencies do.
So what's the statement? Not much, and then a whole lot. The FTC isn't changing its position on consumer protection. It's not requiring any new framework or disclosures. It simply reaffirms the law it already enforces: it's illegal to deceive consumers. It's proposing a statement that ensures that people conducting commerce in the U.S. know that it does apply to AI. When an "large-language models to particular purposes, or a not-yet-developed superintelligence," makes a claim, the company behind it can't hide behind "the model said it." If the system states something as fact, that fact has to be true; if it's really an opinion or a leaning, that has to be disclosed as such.
Here's why that's a whole lot. It means that if an AI product tells the world something false about you — states it flatly, as fact — the company that built the AI can be on the hook for it. Not the model. The company.
That's not hypothetical. A few weeks ago I reported on a case in Germany (link), where a court found Google liable for false claims in its AI Overviews — the summaries had tied two publishers to scams and shady business practices that didn't hold up. (The ruling is a temporary injunction and isn't final yet, so this is still moving.) The court's logic is the interesting part: it said the AI Overview isn't a list of search results, it's Google's own content — Google's own words. That reasoning isn't binding in the U.S., but it's the same idea the FTC seems to be leaning on: the AI's output belongs to the company that made it.
I'm no lawyer, and this isn't legal advice, but in theory that opens a door. If an LLM publishes something about you that's a verifiable fact and it's false, the company could be liable.
The messier question is this: "What about an opinion the AI presents without disclosing that it's an opinion?" Does failing to disclose that make the company liable too? I don't know. If someone can ask a lawyer, I'd love to hear the answer.
The whole statement reads like an attempt to draw a line: whatever new AI laws or executive orders come next, they shouldn't be read to quietly roll back the consumer protections already on the books.
So: if you consume products in the U.S., you get a say here. Go exercise your right: [https://www.regulations.gov/docket/FTC-2026-0859]. Eighty-four people already have. There should be a lot more.
This has a been A Geeky Production
