Authors vs. Artificial Intelligence: Anthropic's landmark deal that may redefine intellectual property

The recent settlement reached by Anthropic, developer of the Claude chatbot, with a group of authors and publishers in the US marks a milestone in the relationship between content creators and artificial intelligence companies.

Anthropic, one of the leaders in artificial intelligence and creator of the Claude chatbot, agreed to pay $1.5 billion, ending a lawsuit in which it was accused of having trained its models with content from literary works without having sought permission from the copyright holders, as required under current law.

While the settlement does not set a judicial precedent as it was settled out of court, it could set a benchmark for how the technology industry should approach the use of copyrighted works in the age of AI. It is estimated that approximately 40 such lawsuits are currently underway.

In another similar case, US courts in the 2000s ruled that file-sharing services such as Napster and Grokster infringed the rights of copyright owners by allowing songs, movies and other copyrighted materials to be shared for free on the Internet.

The lawsuit against Anthropic presents two crucial scenarios:

• The fair use of protected works for the training of AI tools: The court found this type of use to be compliant with the law, being transformative and covered by the fair use doctrine.
• The use of protected content obtained from illegitimate sources: The Court qualified this as direct copyright infringement.

This nuance is essential as it considers that the borderline between innovation and infringement is not in the technology itself, but in the provenance and permissions to use the protected content. Large companies need an enormous amount of data to train and refine their AI systems, and much of this data is protected by copyright, which is where the conflict with authors, publishers, musicians and other artists who want to protect their creations arises.

Faced with this approach, Anthropic opted to agree to pay the rights holders as otherwise the fines set for those who infringe copyright knowingly are much higher. This payment covers only pirated works, not fair use.

Although the regulatory approach to artificial intelligence differs between the US and Europe, the direction of the debate is the same. While the US administration has prioritised fostering innovation and global technological leadership through more flexible regulation, the EU has already passed the first Artificial Intelligence Act, aimed at reducing risks and setting clear limits to the development of this technology.

In this context, the message of the Anthropic case is clear: in the age of AI, respecting copyright is not only an ethical obligation, but a strategic business necessity. Companies developing artificial intelligence systems should incorporate strict legal content acquisition policies from the outset, with specific budgets earmarked for licensing and usage rights, thus avoiding future litigation and strengthening their corporate reputation.

Javier Sabido Manager of the Intellectual Property and TMT Area of ECIJA Chile