A naukri.com initiative
TechCrunch
18h
239
Image Credit: TechCrunch
Court filings show Meta staffers discussed using copyrighted content for AI training
- Meta employees have discussed using copyrighted works for training AI models, according to court documents.
- The dispute in the case Kadrey v. Meta revolves around Meta's claim of 'fair use' for training AI models on IP-protected works.
- Internal work chats among Meta staffers reveal discussions on acquiring copyrighted books for training AI models.
- Meta employees debated acquiring e-books at retail prices instead of pursuing licensing deals with book publishers.
- There were mentions of using publicly available data for model training while being less conservative with approvals.
- Discussions included the consideration of using Libgen, a site known for providing access to copyrighted works.
- Meta's AI team tuned models to steer clear of IP risky prompts and avoid disclosing the use of certain datasets.
- Revelations suggest Meta may have scraped Reddit data for model training and adjusted decisions on training sets.
- Plaintiffs in the case allege Meta cross-referenced pirated books with licensed ones to assess licensing agreements.
- Meta's legal defense team has been strengthened, indicating the company's high stakes in the legal battle.
Read Full Article
14 Likes
For uninterrupted reading, download the app