The Justice Department says training AI on copyrighted content isn’t infringement. The Seattle Times and Newsday say it absolutely is, and they’ve filed a federal lawsuit to prove it. Both of these things are true simultaneously, and if you build bots for a living like I do, that contradiction should keep you up at night.
On September 4, 2026, The Seattle Times Co. and Newsday filed a federal copyright and trademark complaint against OpenAI and Microsoft in the U.S. District Court for the Southern District of New York. Their argument is straightforward: these companies used journalism content without permission to train AI models, and doing so harms the business of journalism. They join a growing roster of publishers who have taken the same legal route.
Why This Matters to Anyone Building on Top of AI APIs
I spend my days wiring up bots — connecting language models to workflows, building retrieval-augmented generation pipelines, and helping teams at ai7bot.com ship actual working architectures. Most of us in this space treat the foundation models as a given. We call an API, get a response, and build our logic around it. We rarely stop to ask: where did the knowledge inside that model come from?
These lawsuits force that question into the open. If courts eventually rule that training on copyrighted material constitutes infringement, the models we depend on could face licensing costs, content restrictions, or even court-ordered retraining. That ripples all the way down to every bot, every agent, every automated workflow sitting on top of those APIs.
As a bot builder, I can’t control what data went into GPT or any other model. But I can — and should — think about what happens if the legal ground beneath those models shifts.
Two Sides, Both Making Credible Points
The publishers’ case is intuitive. Newsrooms invest enormous resources in reporting, editing, and fact-checking. When an AI model absorbs that work and then generates answers that compete with the original source, the publisher loses traffic, loses subscribers, and loses revenue. The Seattle Times and Newsday argue this dynamic is actively harming journalism businesses.
On the other side, the Justice Department has sided with tech companies, arguing that AI training does not constitute copyright infringement. The legal theory here leans on fair use — the idea that transforming copyrighted material into something new (like a statistical model) is fundamentally different from copying and redistributing it.
Both arguments have weight. And that’s exactly what makes this messy.
What I’m Doing Differently in My Own Bot Architecture
I’m not a lawyer, and I’m not going to pretend I know how these cases will resolve. But I am someone who ships production bots, and I’ve started making practical adjustments:
- Attribution layers: When my bots surface information that clearly originated from a specific publication or source, I’m building in citation mechanisms. It’s not legally required yet, but it’s good practice and it future-proofs the architecture.
- RAG with licensed or open data: For retrieval-augmented generation pipelines, I’m increasingly selective about the document corpora I connect. Using openly licensed datasets or content my clients actually own reduces exposure to downstream copyright disputes.
- Monitoring model provider terms: OpenAI, Microsoft, and others update their terms of service regularly. If a major ruling changes the data composition of a model, there could be behavioral shifts. I want to catch those early.
- Keeping humans in the loop: For bots that summarize or rewrite news content, I’m adding human review steps. Automated summarization of copyrighted articles is precisely the kind of behavior these lawsuits target.
A Bigger Pattern Forming
The Seattle Times and Newsday lawsuit isn’t an isolated event. It’s part of a pattern — a steady accumulation of legal pressure from content creators who feel their work has been appropriated without compensation. Each new case adds pressure on the courts to draw a clear line, and each new case raises the stakes for anyone building on top of these models.
If you’re a bot builder reading this, I’d encourage you to follow these cases closely. Not because you need a law degree, but because the tools you build with are shaped by these outcomes. The APIs you call tomorrow may behave differently than the ones you called yesterday, and the reason might be a judge’s ruling rather than an engineer’s commit.
We build smart systems on foundations we don’t fully control. That’s always been true. But the legal battles between publishers and AI companies are making it impossible to ignore. The best thing we can do is build with awareness, design for adaptability, and take the concerns of content creators seriously — because the bots we build are only as solid as the ground they stand on.
🕒 Published: