The Justice Department Says AI Can Train on Your Website. Your Output Is Still Your Problem.

The question I get more than almost any other, in every room, is some version of "am I going to get sued for this." Usually it's an owner who's been using ChatGPT to write product descriptions for eight months and just now started wondering if that was a bad idea. A woman at a session in Maumee asked me flat out if she needed to take her whole About page down.

Something happened this week that changes the answer a little, and I want to be careful about which part it changes.

The Justice Department filed a brief Tuesday in federal court in Manhattan backing OpenAI in the case the New York Times and a group of other papers brought against them. The government's position is that training a large language model on copyrighted material is generally fair use. They called the training "extraordinarily" transformative and said the public benefits outweigh any competitive harm to the people whose work got used. That's the first time our government has said anything on the record in this whole pile of AI copyright cases, and there are a lot of them now. The Times put out a statement saying the administration is siding with a handful of trillion dollar companies against the people whose work got taken.

Two things before anybody gets excited. A brief isn't a ruling. The judge can read it and do whatever she wants. But when the United States shows up and tells a court which way to go, that carries weight, and every other AI case in the country is going to cite this.

Your content is training data and that fight is basically over

Your website copy is in there. So are your photos, your blog posts, the descriptions you wrote for every service you offer, and every Google review anybody ever left you. It got scraped years ago and nobody asked. If you were holding out hope that some court was going to make that stop, this week was a bad week for that hope.

I don't love it either. But I'd rather plan around what's actually true. Anything you publish publicly is going to get read by a machine and turned into somebody's model, and the practical move is to stop treating your website as the thing that makes you different. What makes you different is that you show up, you answer the phone, and you know the guy. None of that scrapes.

The part that can actually bite you

Here's where owners get it backwards. The fair use argument is about training, meaning what OpenAI did to build the thing. It says nothing about what comes out the other end.

If ChatGPT hands you a paragraph that's close to somebody's copyrighted page and you paste it onto your site, you published that. Not OpenAI. The government's brief doesn't help you at all in that situation, and neither does the terms of service you never read. Small businesses have already gotten demand letters over exactly this, usually for a chunk of copy or a photo that came out of a tool and went straight onto a live page.

So three habits. 1: Before anything AI wrote goes live, take one distinctive sentence out of it and Google it in quotes. Takes ten seconds and it catches the copy jobs. 2: Stop asking it to write "in the style of" a named competitor or a named author. That's how you get output that lands close to the original, and it looks awful if it ever gets pulled up. 3: Images are a separate fight with separate cases and this brief doesn't cover them, so don't assume an AI generated photo of a product is clean just because your text is.

My honest read is that this makes AI tools safer to use, not riskier, as long as you own what you publish. Keep using them. Just don't treat the output like it came pre-cleared, because it didn't, and the company that made the tool has already told a federal court that the legal risk lives with the person who hits publish.