AI companies like OpenAI are reportedly willing to pay between $1 million and $5 million annually to license copyrighted news articles for training their AI models. This revelation provides insight into the potential costs AI companies are ready to bear for licensed material. Apple is also reportedly offering at least $50 million over a multiyear period for data to train its AI.
These figures are comparable to some non-AI licensing deals, such as Meta’s offer of up to $3 million a year to license news stories for the now-discontinued Facebook News tab. Google, under legal pressure, agreed to pay Canadian publishers a total of $100 million annually for linking to their articles.
AI models are typically trained on information from the internet, with data often sourced from specific datasets or web crawlers. However, this practice is facing challenges as some companies block access to their data and others argue that training on their data constitutes copyright infringement.
To avoid these issues, AI companies are increasingly partnering with publishers. For instance, OpenAI has signed deals with Axel Springer and The Associated Press to license stories for training models and developing news gathering technology. Google is also reportedly demonstrating an AI tool, Genesis, that generates news stories from facts, to executives from major news organizations.
Read more at The Verge…