Online content has become part of the infrastructure that artificial intelligence companies rely on to develop their models and provide answers to users' questions, but this use raises controversy over the boundaries between making information available to the public and having the right to copy or use it commercially.
Hot topics
The debate is not only about the ability of technology companies to access web pages, but also about the rights of content owners, mechanisms for compensating them, and the extent to which new services affect the digital journalism economy.
The development of artificial intelligence models relies on large datasets, both textual and otherwise, which are processed to extract patterns and relationships that help the models generate answers. One method used to collect internet content is through automated web crawlers that move between pages and links, extracting available text and data before the material undergoes cleaning, organization, and processing.
In its official explanation of its data approach, OpenAI states that its model development sources include publicly available online information, data obtained through partnerships with other entities, and information provided by users or trainers under relevant terms. This diversity of sources means that data collection is not necessarily the same across all companies or at all stages of model development.
Is the public content available for training without permission?
The answer varies depending on the applicable law, the nature of the material, its intended use, and the surrounding circumstances. In the United States , some uses fall under the so-called fair use doctrine, which requires a legal assessment that considers several factors, including the purpose and nature of the use, the nature of the protected work, the size of the part used, and the potential impact of the use on the market for the original work.
The U.S. Copyright Office devoted part of its official study of artificial intelligence to the issue of training models on copyrighted material, indicating that the issue involves legal and policy questions that are not simply a matter of content being publicly available. This does not mean that every use of training is illegal, or that all such uses are automatically covered by the principle of "fair use," as the judgment depends on the details of each case and the applicable law.
The issue becomes even more complex when it comes to paid journalistic articles, images, and archives, or when AI systems use copyrighted material to provide summaries or answers that might eliminate the need for some users to visit the original source. Here, the legality of data collection intersects with the question of economic competition and the impact of its use on the value of journalistic work.
Tags:
technology