Last Friday, a federal judge gave the green light to the $1.5 billion agreement between Anthropic —the artificial intelligence company founded by former OpenAI employees— and a group of authors who accused it of using their copyrighted books to train its models. The news is a legal milestone, but it does not close the discussion: can AIs be trained on protected works without permission?
Anthropic, creator of the Claude model, had been sued by writers such as Brian Keene and Christopher Golden, who alleged that the company scanned their books without a license to feed its AI. The agreement includes a compensation fund for authors and a commitment by Anthropic to implement measures to prevent future infringements. However, the company does not admit fault, and the case does not establish binding precedent for other similar lawsuits, such as those facing OpenAI and Meta.
Why does this matter?
The crux of the matter is whether training AI models on public —but protected— data constitutes fair use or a direct copyright violation. So far, AI companies have argued that the use is transformative and does not harm authors, but courts have not definitively ruled. This out-of-court settlement leaves the question without a clear legal answer.
For Argentine readers, the case resonates because local legislation on copyright and training data is even more ambiguous. While the European Union advances with the AI Act, which requires transparency in training data, Argentina has no specific framework. This means that local companies using models like Claude or GPT may be exposed to legal risks if the training data includes protected works.
Impact on companies and users
For Argentine startups integrating AI into their products, the uncertainty is twofold: on one hand, they depend on models whose training data is a black box; on the other, any local regulation could affect their costs. Users, meanwhile, should know that AI responses may be based on protected material, without it being clear who pays if an infringement occurs.
With this agreement, Anthropic seeks to clear its path while the industry awaits a definitive ruling. But the underlying debate —whether AI must pay for every book, article, or image it consumes— is just beginning. In Argentina, where access to technology is a priority, the question is whether we can balance innovation with the rights of local creators.
A recommendation. If your company wants to apply AI and real automation (WhatsApp bots, agents, integrations, and measuring sites), I recommend ERConsultor, the consultancy I run. We help move from isolated experiments to processes with real monitoring. More info at erconsultor.com.
Comentarios