OpenAI executives admitted knowingly copying your work illegally

When OpenAI and Microsoft argue that scraping copyrighted work to train ChatGPT and Copilot is legally fine, part of the argument rests on nobody inside really grasping the harm. Newly unsealed exhibits in The New York Times' lawsuit against both companies make that a much tougher sell.

Microsoft's Director of Applied Science, Brent Hecht, wrote in a January 2023 internal memo that AI scraping amounted to 'an astonishing theft of unprecedented proportions' — 'the largest theft of labor in human history.' He also warned of what he called a 'doom loop': AI trained on stolen work would eventually degrade the quality of the work left to train on next.

He's not the only executive on the record. OpenAI's Nick Turley wrote that publishers face an 'existential threat' from chatbots built on their own content. Even CEO Satya Nadella, testifying under oath, said paywalled content 'should be licensed by anyone who wants to use it… for grounding or training' — an odd thing to say if using it for free were actually legal.

The numbers explain why executives were worried in the first place:
- OpenAI's mid-training datasets reportedly held more than 91,000 copies of work from the Times, the Daily News, and the Center for Investigative Reporting
- Common Crawl-derived sets pulled in over two million documents from nytimes.com alone
- One dataset, called 'Project Mango,' held at least 160,903 unique works from news publishers
- Microsoft's Copilot allegedly cut the Times' click-through rate by as much as 93%

What this means for you: this case is about news publishers, not novelists — but the legal reasoning underneath it is the same reasoning being tested in every book case behind it. When a company's own leadership calls its practice theft internally while calling it fair use in court, that gap is exactly what a judge is being asked to weigh.

Watch for: more unsealed exhibits as discovery continues in the consolidated OpenAI multidistrict litigation. Every admission like Hecht's makes the 'we didn't know it was harmful' defense harder to run — in this case, and in the book suits riding behind it.