Archive · 16 Jun 2025
Meta's Llama 3.1 AI Model Reproduces Copyrighted Books
Imagine teaching a computer to read books.
Meta built a computer called Llama 3.1 that is designed to do just that.
But it looks like this computer has accidentally memorized entire parts of books, including 'Harry Potter.'
This is like the computer remembering whole pages instead of just understanding the story.
Scientists found that Llama 3.1 can perfectly copy sections from books, which is a problem because the books are protected by copyright.
Other companies are already being sued for similar issues, so Meta might face legal trouble too.
Meta's Llama 3.1 AI model is under scrutiny for reproducing copyrighted book text.
A study found the model memorized approximately 42% of 'Harry Potter and the Sorcerer's Stone'.
Llama 3.1, released in July 2024, showed significantly increased verbatim retention compared to earlier models.
The model also reproduced excerpts from works such as 'The Hobbit' and '1984'.
The findings raise legal concerns for Meta amid growing copyright infringement lawsuits.
- Who
- Meta developed the Llama 3.1 AI model. Researchers from Stanford, Cornell, and West Virginia University conducted the study.
- What
- Meta's Llama 3.1 AI model is reproducing copyrighted text from books.
- When
- The study was conducted recently; Llama 3.1 was released in July 2024.
- Where
- N/A
- Why
- The model's ability to reproduce copyrighted content raises concerns about copyright infringement and potential legal repercussions.
Potential Copyright Holders
Meta
Copyright Infringement
Potential Copyright Holders
Copyright holders are concerned about unauthorized reproduction and distribution of their works by AI models.
Meta
N/A
- Model Name
- Llama 3.1
- Developer
- Meta
- Release Date
- July 2024
- Memorization Rate (Harry Potter)
- 42%
- Other Copyrighted Works
- The Hobbit, 1984
- Legal Action Example
- The New York Times lawsuit against OpenAI and Microsoft