Lawyers for The New York Times and 11 other major publications are using Microsoft’s and OpenAI’s own internal messages, unsealed this week in a summary judgment brief, against them. In one instance, the documents quoted a Microsoft director who called the companies’ artificial intelligence training “an astonishing theft of unprecedented proportions.”
The case stems from OpenAI and Microsoft’s use of the publications’ articles to train and power their AI products, including ChatGPT and Copilot. Both companies say that their use is legal under the fair use rules. But the plaintiffs use the AI companies’ own internal messages as their explicit legal argument for why fair use doesn’t apply here.
Download the Straight Arrow app today to get the stories that matter free from manipulation, bias or agenda.™
Point phone camera here
Fair use relies on a four-factor test, and plaintiffs argue that the first factor, purpose and character of the use, hinges on whether the copying is “transformative” or “substitutive.” Since the Copyright Act of 1976, the Supreme Court has found commercial copying to be fair use only once, according to the brief. The plaintiffs argue the companies admitted in their internal messages that ChatGPT and Copilot do neither, instead replacing the original content the chatbots trained on.
What this case could decide
If OpenAI and Microsoft win, the copyright infringement claims collapse, and publishers could realistically find compensation for AI training only through legislation.
If the publishers win, the damages could be massive. Applying the statutory range for willful infringement, up to $150,000 per work, to the roughly 10.8 million articles the plaintiffs say are at issue would put theoretical damages at about $1.6 trillion. That’s a statutory ceiling, not a figure either side has put forward. A win could also revive a request from the plaintiffs’ original complaint that the companies destroy the models trained on their content.
Brent Hecht, the Microsoft employee the publications cited in their brief, was more descriptive in his internal messages, saying people would soon believe AI models were “hoovering up” other works, the brief stated. He also admitted that people or companies never intended for their content “to be used in this fashion, nor are they compensated for its use.”
AI companies push back
Microsoft has pushed back on the publications’ framing. Microsoft spokesperson Alex Haurek told The Times that Hecht’s messages didn’t represent the company’s views. He pointed to the company’s court filings, which argue “Copilot is not a substitute for publishers’ journalism.”
In a separate filing, Microsoft described Hecht as a research academic who was also affiliated with Northwestern University, The Times wrote, and was “not a decision maker.” Rather, the company hired him “to present divergent and asymmetric perspectives.”
OpenAI didn’t respond to The Times’ request to comment.
The pink slime problem
The brief also uses pink slime to bolster its market-harm argument. The news groups say the AI tools these companies develop are driving a wave of new pink slime news, or unreliable low-quality content plagiarized from legitimate news sources. The plaintiffs argue AI has made it “cheaper, easier to make, and harder to detect.”
The brief uses OpenAI’s published pricing, stating it would cost less than $7,000 to generate one million articles averaging 500 words each without any human involvement.
This isn’t a hypothetical either. The brief points to real-world incidents where this is already playing out. One example was Prism News, a network of more than 200 AI-generated sites — including 84 local news sites and other hobby publications — run by just four employees. It has since shut down, saying that the “experiment” was finished. But during the site’s time, it published news stories that misrepresented situations, including a story that resurfaced a death from a previous year’s flood and presented it as current, leading some residents to believe a neighbor had just died.
What’s next?
The legal fight is far from over, with OpenAI, Microsoft and the publishers each filing competing summary judgment motions in early September asking the judge to rule on liability and fair use before any trial. No trial date has been set.
Days before the filings, the Justice Department filed a statement backing OpenAI’s position, arguing AI training is fair use and calling the publishers’ market-dilution theory “deeply flawed.”
The case might also face a pause. The judge has ordered The Times to explain why its case shouldn’t be stayed while the other motions are decided, with the defendants’ response due Friday, according to Deadline.
Round out your reading
- Lawyer says 57 Cornell frat members were in the group chat. Could their silence be a crime?
- Florida county discovers mysterious Flock cameras with no obvious owner.
- Off Script: They won their primaries on one message. Now it’s disappearing.
- Some question whether a popular affordable housing policy is making the nation’s home shortage worse.
- Kelvin wave incoming: Why the US should care.
