Court filings show Microsoft and OpenAI admitted content theft
On September 17, 2026, previously redacted court documents were unsealed in the long-running lawsuit between The New York Times and OpenAI and Microsoft.
According to these documents, a Microsoft executive privately described AI training on outside content as 'an astonishing theft of unprecedented proportions' and 'the largest theft of labor in human history.'
The filings state that OpenAI's training datasets contained more than 91,000 copies of articles from the New York Times and other publishers, with one additional dataset reportedly including over two million documents from nytimes.com alone.
Microsoft CEO Satya Nadella testified under oath, according to the filing, that click-through rates to New York Times articles via Bing dropped by as much as 93 percent after AI systems began using the paper's content.
Internal documents describe a 'doom loop,' in which AI products draw on publishers' content while simultaneously diverting readers and revenue away from them, which one cited Microsoft document said threatens the economic stability of its own content suppliers.
According to the filings, OpenAI itself described its products as an 'existential threat' to news publishers, and one employee is said to have testified that users rarely click through to original sources even when links are shown prominently.
It remains unclear how the court will legally weigh these statements, since earlier rulings have often found AI training to qualify as 'fair use,' and the underlying evidence documents remain sealed.
WHY IT MATTERS
The filings show, through the companies' own internal statements, how heavily AI systems rely on copyrighted work from journalists and creators and how much this can threaten their livelihoods.
WHAT YOU CAN DO NOW
- Visit original sources directly Click through to news sites or websites instead of relying only on a chatbot's summary. This helps ensure creators still get visits and revenue.
- Support journalism you value If an article matters to you, subscribe to or visit the source directly. This helps keep newsrooms funded.
- Double-check AI answers For important topics, check whether an AI answer is based on a real, verifiable source. When in doubt, look up the original source yourself.
- Think about where content comes from When an AI tool generates text or images, consider where its training data might have come from. This builds awareness of original creators' rights.