The New York Times sues OpenAI and Microsoft

The New York Times filed suit against OpenAI and Microsoft in the Southern District of New York, alleging copyright infringement in the use of millions of its articles as training data. It was the first major American news organisation to sue an AI developer, and it sought both damages and the destruction of models trained on its work.

The complaint’s distinguishing feature was its evidence. Rather than arguing only that training was infringement, it included roughly a hundred examples in which GPT-4 was said to reproduce Times articles at length and near-verbatim, including from behind the paywall, and further examples in which Bing Chat attributed fabricated statements to the paper. The regurgitation exhibits made the case harder to dispose of on general fair-use grounds.

OpenAI responded publicly in January 2024, arguing that training on publicly available material is fair use, that regurgitation is a rare bug it works to eliminate, and that the Times had produced its examples through prompts specifically engineered to elicit them. It also disclosed that talks over a licensing deal had been under way.

The suit arrived alongside a growing set: authors’ class actions against OpenAI and Meta, Getty against Stability, visual artists against several image generators, and later music publishers against Anthropic and record labels against Suno and Udio. Several defendants settled or licensed instead; News Corp, Axel Springer, the Financial Times and the Associated Press all signed content deals with OpenAI in the following eighteen months, which made the Times’s refusal the conspicuous exception.

The underlying question — whether training a commercial model on copyrighted text is transformative fair use — was the single largest legal uncertainty hanging over the industry for the rest of the period.