Home / Technology / AI Training Data Dispute Escalates to Sanctions
AI Training Data Dispute Escalates to Sanctions
10 Jul
Summary
- Publishers seek sanctions against OpenAI for withholding evidence.
- Allegations concern using copyrighted works to train AI models.
- AI overviews are linked to declining digital media traffic.
Multiple news organizations have accused OpenAI of deliberately withholding evidence related to how its artificial intelligence models are trained. This accusation comes as part of a new legal motion connected to ongoing copyright lawsuits filed against the company. The motion was submitted by 17 publishers, asserting that OpenAI has engaged in "obstruction" by failing to produce necessary evidence such as datasets and output logs.
This legal action stems from earlier lawsuits, including one initiated by The New York Times in 2023 against OpenAI and Microsoft, alleging the use of millions of news articles for AI development. Ziff Davis also sued OpenAI in 2025, claiming its copyrighted content was extensively used and monetized without consent. OpenAI maintains that its AI training practices constitute fair use and denies the allegations, with a spokesperson stating the claims are "blatantly false."
The broader implications of these lawsuits extend to the significant traffic declines observed across digital media outlets. Generative AI, particularly AI overviews, is cited as a major factor siphoning readership and advertising revenue, with some small publishers reporting traffic drops as high as 60%. Courts are currently grappling with defining the boundaries between fair use and copyright infringement in the rapidly evolving field of AI.
