DOJ Backs OpenAI in Copyright Lawsuit With New York Times
Left 0%
Center 0%
Right 100%
0 left · 0 center · 2 right
What happened
The U.S. Department of Justice filed a statement of interest in Manhattan federal court this week supporting OpenAI in a copyright lawsuit The New York Times brought in December 2023 against OpenAI and Microsoft, alleging they used millions of Times articles to train AI products that compete with the newspaper. The DOJ argued that training large language models on copyrighted texts does not violate copyright law, that a ruling for the Times would threaten U.S. national security by advantaging foreign adversaries not bound by similar limits, and that mandatory licensing would leave only the largest technology companies able to compete in the market for such models. Microsoft separately asked the judge for summary judgment, citing analysis of 8.2 million Copilot chat logs that it said showed regurgitation of news and book text was rare. A federal judge had previously declined to dismiss the core copyright-infringement claims while narrowing the case, and related publisher and author suits were consolidated before one judge. The brief was the first time the federal government has formally intervened in litigation over whether AI training on copyrighted material is lawful.
Omitted — what each side leaves out
Unpacked
Left-leaning coverage never reports the Justice Department's reasons for backing OpenAI in the Times suit, noting only that a statement of interest was filed. That mention sits in a parenthetical in The Verge, which otherwise covers Microsoft Copilot logs. Right-leaning coverage is simply more complete on the event, reporting the government's national-security claim and its warning that licensing would leave only the largest firms able to compete.
Left-leaning coverage leads with Microsoft's claim that Copilot rarely reproduces news text and puts that rarity in the headline. Right-leaning coverage leads with the department siding with OpenAI and does not report those discovery results at all.
Would the department's test of "significant substitutive competition" be met by the regurgitation rates Microsoft filed the same week?
Bottom line
A reader of The Verge would not know the DOJ told the court that limits on training would "threaten national security," a line Breitbart prints from the same brief.
The receipts — 5 facts one side's coverage lacks
- The Justice Department filed a brief arguing that using copyrighted content to train AI models does not violate the law.
- Fewer than 1 percent of more than 8 million Copilot chat logs regurgitated at least 16 words.
- The administration's AI framework said Congress should consider licensing systems for rights holders, and the DOJ brief does not fully align with that framework.
- Only 10 of 212 books evaluated had any matches in the Copilot conversations Microsoft produced.
- A federal judge last year declined to dismiss the core copyright infringement claims in the Times suit against OpenAI.
The Left View
Left-leaning coverage focused on Microsoft's claim that Copilot almost never substitutes for news or books, and on the Times' insistence that the products were built on stolen journalism. In those Copilot logs — selected, Microsoft said, because they hit keywords implicating the news plaintiffs' sites — Microsoft reported that fewer than 1 percent regurgitated at least 16 words, that 59,545 logs had at least 16 words in common with news content used to ground the model, that a Center for Investigative Reporting expert found 51 instances of "substantial overlap" with CIR work, and that an expert in the authors' suit found only 24 responses containing at least 30 matching words, with matches in only 10 of 212 books evaluated. Microsoft argued those figures support fair use because the systems are used for significantly different purposes than the originals and that occasional reproduction "hardly undermines the transformative purpose of LLM training." Times lead counsel Ian Crosby said the discovery materials "lead to only one conclusion: Microsoft and OpenAI stole from The New York Times to make commercial products that substitute for its journalism, threaten its business, and undermine its industry."
The Right View
Right-leaning coverage presented the DOJ brief as a national-security and competition argument that training is fair use, while recording the Times' and publishers' objections and a possible split with White House policy. The department wrote that the United States has "a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law," that rules making a robust U.S. AI industry "significantly more difficult" "threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered," and that it would be "problematic—and legally incorrect—to impose broad copyright liability that would generally render training of AI models impermissible without licensing." It also asked the court not to adopt reasoning from a separate authors' suit against Meta that, the DOJ said, wrongly treated training and outputs as one continuous act and accepted market harm from the mere potential of competing AI-generated books. A Times spokesperson said the administration is "siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole" and that "AI companies simply need to pay fairly for the content that makes their products possible, as copyright law requires." A former FCC adviser, Evan Swarztrauber, and the News/Media Alliance noted the brief's harder anti-licensing posture versus the administration's March AI framework, which stated that training does not violate copyright law, left the question to the courts, and asked Congress to consider collective licensing systems without deciding when licensing is required.
Our Take (balanced)
The strongest argument on the technology-company side is Microsoft's log analysis showing that even keyword-selected Copilot chats seldom reproduce substantial stretches of news or book text, which it and the DOJ treat as evidence that training is a transformative use rather than a market substitute, reinforced by the DOJ's claim that treating training as infringement would advantage "foreign adversaries who are not so encumbered" and confine the market to firms that can pay licenses. The strongest argument on the publishers' side is that unlicensed ingestion of copyrighted articles to build commercial products that "substitute for its journalism" is theft under copyright law regardless of how often full passages reappear, and that companies "simply need to pay fairly for the content that makes their products possible, as copyright law requires." The central unresolved tension is whether fair use turns on the claimed rarity of regurgitated outputs and the government's national-security and competition concerns, or on the unlicensed taking of copyrighted works as training inputs and the claimed substitutive threat to the market for human-created news and books.
3 sources
More stories
The week's bottom lines, in your inbox
One email a week: the five stories that mattered and what they actually mean. Free.