OpenAI, Google and Microsoft broker deals with major news outlets – are music companies next?

Credit: Gonzalo Poblete

OpenAI, Microsoft and Google mediate data contracts with news agencies. Are music companies next?

Reports from major news outlets like the Financial Times suggest these companies are meeting with news executives over potential copyright issues in the creation of AI content. Large language models like ChatGPT and generative AI models like Midjourney and Stable Diffusion are trained on existing content. Stable Diffusion is facing a lawsuit from Getty Images over the Getty watermark appearing in AI-generated images – meaning Getty Images were part of the data set to train Stable Diffusion.

theft reports that executives from News Corp, The New York Times, The Guardian and Axel Springer have all discussed a subscription fee for content used to develop artificial intelligence models. An agreement with these news organizations could lay the groundwork for similar copyright issues in generative AI. So what does this pricing model look like? It’s too early to tell, but early reports suggest that accessing content costs between $5 million and $20 million a year.

Among other things, the presentations will discuss the creation of a “quantitative model” for accessing data – similar to what is already used in the music industry. Radio stations, nightclubs and streaming services all pay record companies when a track is played, and that’s exactly what news outlets want to see from AI companies. Every time a piece of information is accessed, that owner gets paid. But to establish such a model, said companies would need to disclose the use of media content in their current operations – which these companies are unwilling to do.

Sam Altman, CEO of OpenAI, even suggested that his company would cease operations in the EU if the bloc enacted regulations governing its operations. The EU AI law would force companies like OpenAI to disclose copyrighted material used to train their systems.

“If OpenAI cannot meet the basic requirements for data management, transparency and security, then their systems are not suitable for the European market,” says Dutch MEP Kim van Sparrentak, who helped draft the current EU AI law stands.