The New York Times sues OpenAI and Microsoft over copyright

If someone took decades of your life's work, used it to build a multi-billion dollar tool, and that tool started reciting your work word-for-word behind your own paywall, you would probably call your lawyers too.

The New York Times officially sued OpenAI and Microsoft for copyright infringement in federal court. The publication claims the two tech giants copied millions of its articles without permission to train large language models like ChatGPT and Copilot. The lawsuit asks for billions in statutory damages, an injunction, and the outright destruction of any datasets containing Times content.

Reciting news verbatim and breaking paywalls

Unlike earlier, broader lawsuits from authors and artists, the Times came with receipts.

According to the complaint, in the Common Crawl dataset used to train earlier versions of GPT, there were roughly 16 million unique records from Times-owned domains. That made the Times the third-largest single source in the entire dataset, trailing only Wikipedia and a massive US patent database.

The real punch in the lawsuit is what happens on the user end. The Times showed that ChatGPT and Microsoft Copilot do not just learn general writing styles; they can actually spit out entire paragraphs of copyrighted, paywalled reporting verbatim.

On top of that, the lawsuit complains about hallucinations, noting instances where AI tools invented false information and attributed it directly to Times reporting, damaging the paper's brand while siphoning away subscription, licensing, and ad revenue.

A broken deal and a divided media world

This legal blowup did not happen overnight. The Times reportedly spent months negotiating behind closed doors with both OpenAI and Microsoft to establish a paid licensing framework, but those talks fell apart.

OpenAI sounded caught off guard. Spokesperson Lindsey Held stated the company was surprised and disappointed, claiming their ongoing discussions had been productive and constructive, and expressed hope for finding a mutually beneficial resolution.

Other media publishers have taken the cash instead of filing lawsuits. Axel Springer, owner of Politico and Business Insider, recently struck a multi-million dollar deal to feed content directly into ChatGPT. The Associated Press also signed a two-year licensing pact allowing OpenAI to train models on its archives.

At the same time, outlets like the BBC, CNN, and Reuters have moved aggressively to block OpenAI web scrapers from hoovering up their sites.

Why it matters

This is the biggest legal test for generative AI to date. It is effectively the Napster moment for large language models, and how the court rules could reshape how AI is built from the ground up.

In our view, the Times is not trying to kill artificial intelligence; this is an aggressive, high-stakes negotiating play. They want to set a clear precedent that premier training data has a steep price tag, and tech companies can no longer treat the open web as an all-you-can-eat free buffet.

If the Times wins or forces a massive settlement, the cost of training frontier AI models will skyrocket overnight. Tech giants with deep pockets like Microsoft and Google can afford multi-million dollar licensing deals. Smaller AI startups and open-source researchers, however, could find themselves completely priced out of clean, high-quality training corpora.

There is also a hidden risk for the publishing industry itself. As several copyright experts have pointed out, if the courts declare that summarizing and learning from public text infringes on copyright, traditional journalism outlets that regularly summarize scoops from competitors might find themselves staring into the exact same legal barrel.