Unsealed files prove AI firms knew copying illegal

For years, OpenAI's defense against author lawsuits has leaned on a simple story: nobody really knew the training data included pirated books, and even if they did, training a model on it is transformative fair use. That story just got a lot harder to tell.

Newly unsealed filings in the Authors Guild's copyright suit against OpenAI and Microsoft show the company's own staff knew — and put it in writing. An internal note from August 2019 says, bluntly: 'We trained GPT-3 on pirated stuff! No sharing that!' Nine months later, OpenAI policy director Jack Clark told colleagues that 'the better we do on GPT-X, the more worried genre fiction authors will become about us substituting for them on Amazon' — and predicted the company would 'likely ignore their concerns and release anyway.'

He wasn't wrong. The filings describe Tarun Gogineni, hired in 2022 to sharpen the model's prose, treating author complaints about stolen datasets as 'acceptable economic disruption' rather than a problem worth solving.

The piracy source at the center of it all is LibGen, the shadow library authors have been fighting for a decade. According to the filings:
- Microsoft knew OpenAI was pulling from LibGen as early as April 2019
- Sam Altman and Dario Amodei reportedly disclosed the LibGen use to Bill Gates and Microsoft CTO Kevin Scott during an early GPT-3 demo
- By 2022, as scrutiny mounted, OpenAI ran an internal effort called 'Project Clear' to scrub the LibGen files, with a VP asking colleagues how to make the data disappear quietly

Why this changes the fight: a fair-use defense gets much weaker once a company can be shown to have known the material was stolen and decided shipping anyway was worth the risk. That's not an accident-of-scraping story. That's a business decision, on paper, with a trail.

What this means for you: if you write genre fiction — romance especially, per Clark's own words — you were the example used internally to justify moving forward regardless. That paper trail is now evidence working in your favor.

Watch for: summary judgment briefing in the Authors Guild case continues through the rest of 2026. If this evidence holds, it strengthens every author case still standing behind it, including any settlement talks that follow.