News outlets demand new rules for AI training data

Several media organizations called for rules protecting copyright in data used to train generative AI models. 

The open letter urged lawmakers around the globe to consider regulations that require transparency into training datasets and consent of rights holders before using data for training. They also request allowing media companies to negotiate with AI model operators, identifying AI-generated content, and mandating AI companies eliminate bias and misinformation in their services. 

Signatories of the letter include Agence France-Presse, the European Pressphoto Agency, the European Publishers’ Council, Gannett, Getty Images, the National Press Photographers Association, the National Writers Union, News Media Alliance, The Associated Press, and The Authors Guild. 

The signatories said foundation models trained using media content disseminate information “without any consideration of, remuneration to, or attribution to the original creators.”

“Such practices undermine the media industry’s core business models, which are predicated on readership and viewership (such as subscriptions), licensing, and advertising,” the letter said. “In addition to violating copyright law, the resulting impact is to meaningfully reduce media diversity and undermine the financial viability of companies to invest in media coverage, further reducing the public’s access to high-quality and trustworthy information.” 

News organizations aren’t the only ones concerned about AI models training on copyrighted material — a practice whose legal status remains untested. The Senate took up the issue in several hearings, and a lawsuit alleging generative AI art platforms Midjourney and Stable Diffusion infringed on artists’ rights is making its way through the court. Comedian Sarah Silverman and two authors sued OpenAI for alleged copyright infringement.

The letter’s signatories noted they believe generative AI may offer significant benefits to organizations and the public while asking to participate in discussions in respecting media companies’ rights. 

Reuters reported that some signatories already have deals to allow AI companies to use their material for training. The Associated Press, for example, allowed OpenAI to license part of its archive and explore using generative AI for news writing.