Humanity wins an important AI copyright case, but remains on the hook to use pirated books

AI For Business


Humanity has won a great legal victory over whether artificial intelligence companies were justified by knitting millions of copyrighted books to train chatbots.

In a ruling that could set a significant precedent for a similar dispute, Judge William Alsap of the U.S. District Court for the Northern District of California on Tuesday said he did not violate US copyright law using books he legally purchased to train Claude, an AI model.

Founded by a former executive of ChATGPT developer Openai, humanity introduced Claude in 2023. Like other generator AI bots, this tool allows users to ask natural language questions and then provide neatly summarized answers using AI trained with millions of books, articles and other materials.

Alsup determined that humanity used copyrighted books to determine that the Language Learning Model (LLM) was “typically transformative” and did not violate the “fair use” doctrine under copyright law.

“Like readers who are aiming to become writers, the LLMS of humanity trained their works to turn hard corners and create something different, rather than racing, recreating or replacing them first,” his decision states.

In contrast, Alsup also discovered that humanity may have broken the law when it downloaded millions of pirated copies individually, and said it would face another trial in December on the issue.

Court documents revealed that human employees have expressed concerns about the legality of using pirate sites to access books. The company later shifted its approach and hired a former Google Executive, who was responsible for Google Books. It is a searchable library of digitalised books that have successfully fought for years.

The author filed a lawsuit

Humanity supported the ruling.

“The court is pleased to recognize that using work to train LLMS (language learning models) is transformative,” a human spokesman told CBS News in an email.

The ruling comes from a lawsuit filed last year by three authors in federal court. After humanity trained Claude with copies of their books, Andrea Burtz, Charles Graeber and Kirk Wallace Johnson sued humanity for allegedly infringing copyright, claiming that the company's practices amounted to “massive theft.”

The author also claimed that humanity “We try to benefit from stripping the human expressions and ingenuity behind each of those works.”

CBS News requested the author for comment but did not respond to Bartz or Wallace Johnson. Graver declined to comment.

Other AI companies are also under attack on the materials they use to build language learning models. For example, the New York Times sued Open AI and Microsoft in 2023, claiming that tech companies used millions of articles to train automated chatbots.

At the same time, some media companies and publishers are seeking compensation by licensing content from companies such as humanity and Openai.

Meta also won a major victory this week after dismissing a lawsuit challenging the methods federal judges used to train artificial intelligence technologies. The incident was brought about by a group of well-known authors, including comedian Sarah Silverman and author Jacqueline Woodson.

I contributed to this report.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *