US government backs OpenAI against The New York Times

The US government backs OpenAI in the copyright fight with The New York Times. On 1 September the Justice Department filed a statement of interest in the federal court in Manhattan, where the case is being heard.

A statement of interest is a letter in which the government tells a judge what it thinks about a case it is not a party to. The judge does not have to follow it. It still matters, because it sets out what the administration wants the law to say.

The government calls model training exceedingly transformative

The filing argues that training a large language model on protected text is fair use. Fair use is the American rule that allows limited reuse of protected work without asking first. According to IPWatchdog, the government calls the training “exceedingly transformative”, because a model turns text into numbers and then writes, translates and answers questions, instead of doing what a news article does.

The filing adds that a win for the newspaper would slow American AI work and help foreign rivals. It also says that if every model had to be licensed, only the largest firms could afford to build one.

Small newsrooms are the argument the filing leans on

One part of the filing is aimed at smaller publishers. It says newsrooms with small budgets need open access to language models to compete, for example to make pictures for an article or to draft an opposing view. Nieman Lab notes that those same publishers rarely get paid for the text used in training, and rarely have the money to sue.

The New York Times answers with a line about trillion-dollar firms

Graham James, a spokesman for the newspaper, said the administration was “siding with a handful of trillion-dollar companies”, in remarks reported by Nieman Lab. Matt Topic, a lawyer for The Intercept, warned of “an unprecedented, uncompensated transfer of IP rights” from newsrooms to technology companies. The case is bigger than one title: book authors, papers owned by Alden Global Capital and non-profit newsrooms are suing too.

Europe is running the same fight with different tools

Regulators here are testing the same question in other ways. The EU now treats ChatGPT like a very large search engine, with the duties that come with that label. In Poland, the telecoms regulator has delayed its ruling on Google paying publishers. And the models are already a media business of their own, with ads inside ChatGPT running at about a billion dollars a year.

For media owners the practical point is narrow. If courts settle that training on news is fair use, a licence deal becomes something an AI company chooses to sign, not something it has to sign. The price of an archive would then be set by goodwill rather than by law.