Editors process Microsoft and OpenAI by scraping content: 'Billions stolen from creators'

Microsoft and OpenAI are being processed by a group of publishers for using millions of copyrighted articles to train their artificial intelligence models without authorization. The action, opened this week, claims that companies have made billions of dollars at the expense of the work of journalists and writers, without paying a penny of compensation. The case may set a crucial precedent for the relationship between generative AI and intellectual property.

What does the suit against Microsoft and OpenAI claim about scraping?

According to information from Bloomberg Law, the collective complaint was filed in a federal court in the United States and involves publishers who claim to have had their content scraped off on a large scale to feed OpenAI language models, such as the GPT-4, and Microsoft integrated services, such as the Copilot. The central charge is that companies violated copyright by reproducing, distributing and publicly displaying articles without licensing. The publishers ask for compensation for damages that can reach $150,000 per work infringed, in addition to a court order that requires the removal of the contents of the training databases.

Why does content scraping affect creators and publishers?

Unauthorized scraping of texts, images and audios to train AI models has become a common practice among big techs, but is rarely questioned in such a comprehensive manner. This process, by involving Microsoft and OpenAI companies, can define clearer rules on the use of public data for commercial artificial intelligence purposes. For journalists, authors and publishers, the decision can mean the recognition that creative work cannot be exploited free of charge — and that the billions of dollars moved by the AI should in part return to those who produced the original content. The case echoes recent discussions on AI regulation, such as the alert made by the Five Eyes alliance on the risks of cyberattacks driven by these technologies — a topic that we address in another analysis.

How do Microsoft and OpenAI justify scraping content?

Until the closing of this edition, neither Microsoft nor OpenAI officially pronounced on the process. In similar previous cases, OpenAI argued that the use of public data fits the concept of fair use (fair use) provided for in American law, especially when the material is transformed into new knowledge by the AI model. However, the defense faces the challenge of proving that the replication of entire parts of articles in ChatGPT or Copilot results does not constitute direct violation. Microsoft, in turn, has invested heavily in the partnership with OpenAI and can try to isolate its responsibility, claiming that scraping was conducted by the model developer. The outcome will depend on technical evidence on how the data were collected and stored.

FAQs about scraping content

What are the publishers asking for in the action?

Editors require compensation for material and moral damages, as well as an injunction that requires Microsoft and OpenAI to delete all copyrighted content from training data sets and pay future royalties for use.

Which publishers are involved in the process?

The action was moved by a group of media publishers, including names such as The Intercept, Raw Story and AlterNet, as reported by Bloomberg Law. Other publishers can join later.

What does that mean for the future of the generative AI?

If the court decides against companies, the decision may force the industry to rethink the way it obtains data for training, encouraging the creation of paid licenses and agreements with copyright holders, which may make the development of new models more difficult.