OpenAI just dropped a blog post showcasing excerpts from 20 amicus briefs filed in support of their fair use defense against The New York Times lawsuit. The framing is aggressive: the Times lawsuit "asks everyone to ignore a half century of legal history" in pursuit of "an undeserved payday."
This isn't a legal brief. It's a public relations document masquerading as a helpful roundup of expert opinion. And honestly? It's well done—which makes it worth dissecting.
The Curation Strategy
OpenAI didn't just publish a bunch of amicus briefs. They excerpted them, organized them, and introduced them with sharp editorial framing. The post opens by positioning fair use as the hero of "decades" of innovation—"medical and scientific breakthroughs" and "new technologies that have made real and positive differences."
Then they pivot immediately to the villain: the Times, asking us to "toss aside" longstanding legal principles. Notice the work that opening paragraph does. It's not defending OpenAI's specific training practices. It's recontextualizing the entire lawsuit as an attack on progress itself.
The amicus excerpts that follow are grouped by filer—accessibility advocates, authors' groups, law professors, tech associations, even the U.S. government. Each excerpt is presented without commentary, letting the authorities speak "in their own words." Clean. Authoritative. Suspiciously one-sided.
What Gets Quoted
The excerpts themselves are fascinating in what they emphasize. Several hit the same talking points:
-
Transformative use: The Computer and Communications Industry Association brief calls AI training "among the most clearly transformative uses a court has been asked to evaluate." Works aren't "copied to be consumed or distributed; they are analyzed to extract statistical patterns."
-
Non-substitution: Authors Alliance argues against the "market dilution theory," saying copyright shouldn't protect against "aggregate competitive pressure from noninfringing outputs." The line being drawn: displacement isn't substitution.
-
Public benefit: Multiple briefs cite medical research, accessibility tools, education. The American Council of the Blind brief name-checks "medical researchers and patients, educators serving students with different learning needs, people with disabilities seeking greater independence."
-
Tool vs. content: Engine and Developers Alliance emphasize that training builds "a general-purpose tool that others can adapt to new ends," not a reproduction mechanism.
These aren't random selections. They're building a coherent narrative arc: AI training is transformative, it creates tools not substitutes, and restricting it would harm vulnerable populations who benefit from the technology.
The Voices That Carry Weight
OpenAI is clearly banking on some sources having more rhetorical punch than others. The United States Government gets its own section header. So does the American Council of the Blind. These aren't just legal arguments—they're coalition-building.
The Authors Alliance brief is particularly clever positioning. Here's an authors' group arguing against the Times' theory of market harm. The excerpt focuses on how copyright "deliberately leaves that competition outside the copyright owner's control" when new works reduce the value of old ones through genre competition rather than substitution.
Kevin Kelly, the author and futurist, gets quoted saying AI offers "an entirely new medium" and warning that a ruling for plaintiffs would "hinder the development of technologies capable of disseminating previously unimaginable quantities of information." It's the same argument, but from a creator's mouth instead of a corporation's.
What's Missing
Here's what you won't find in this post: any acknowledgment that this is a selection. No mention that other amicus briefs might have been filed on the other side. No engagement with the Times' actual arguments beyond the dismissive opening paragraph.
The post also doesn't grapple with any of the harder questions these briefs raise. What is the line between market dilution and substitution when an LLM can generate articles in the style of specific outlets? How do we measure "transformative" purpose when the outputs can be near-substitutes even if the training process is analytical?
The Electronic Frontier Foundation excerpt dismisses the idea that fair use should "tilt...against the developer of a general-purpose tool if that tool might be used to create potentially competing non-infringing works." That "potentially competing" is doing heavy lifting. The whole case turns on whether ChatGPT's outputs do compete with Times articles in ways that matter under copyright law.
The Accessibility Shield
The accessibility framing deserves special attention. Multiple briefs emphasize benefits to people with disabilities, students with different learning needs, and underserved communities. This isn't accidental.
It's much harder to argue against a technology when doing so means arguing against screen readers, educational tools, and medical breakthroughs. The American Council of the Blind brief isn't wrong—fair use does enable accessibility tools, and restricting it would have downstream effects. But that's also not the same as saying OpenAI's specific training practices on Times articles are necessary for those benefits.
The rhetorical move collapses the general principle (fair use enables good things) with the specific case (therefore this particular use must be fair). It's effective advocacy, but it's advocacy.
The Transformative Use Question
The core legal argument across most excerpts is that training is "transformative" because it analyzes text to extract patterns rather than reproducing it for consumption. This tracks with how courts have historically thought about fair use in other contexts—Google Books' text mining, thumbnail images in search results, etc.
But there's a tension the excerpts don't explore. Yes, the training process is analytical. But the outputs can be consumptive. If I ask ChatGPT to summarize a Times article, or write something in the style of a Times columnist, am I using the training data transformatively or substitutively?
The briefs want to separate these cleanly: training is one use (transformative), outputs are separate uses (judge each on its own merits). That's a defensible legal position. It's also one that conveniently insulates training from any downstream market effects.
What This Document Actually Does
This blog post isn't trying to convince a judge. The actual briefs are filed with the court; this is a curated summary published on OpenAI's website. The audience is everyone else: journalists writing about the case, policymakers thinking about AI regulation, and the broader public forming opinions about whether OpenAI is the good guy or the bad guy in this fight.
It's pre-emptive reputation management. By platforming accessibility advocates, authors' groups, law professors, and government briefs, OpenAI is building a permission structure for people to support them without feeling like they're just siding with a tech giant against creators.
The message: this isn't OpenAI vs. the Times. It's innovation, accessibility, and progress vs. a narrow attempt to extend copyright beyond its traditional bounds.
The Meta-Game
What makes this interesting from an AI industry perspective is how much of the debate is happening in public rather than just in court. Both sides are litigating in the court of public opinion alongside the legal case.
The Times published their lawsuit with detailed examples of ChatGPT reproducing their content. OpenAI published rebuttals on their blog, then this compilation of supporting briefs. Each move is as much about narrative control as legal strategy.
The AI industry is watching this case closely because the outcome will shape what's possible for future models. If the court sides with the Times, training on copyrighted content without licenses becomes much riskier. If the court sides with OpenAI, the current free-for-all continues.
That's why you see so many industry groups filing amicus briefs, and why OpenAI is platforming them so prominently. The law here is genuinely uncertain—fair use is inherently fact-specific and context-dependent. Which means the framing work matters enormously.
The Unspoken Trade-Offs
The briefs OpenAI excerpts all emphasize the benefits of permissive fair use: innovation, accessibility, competition, new forms of creativity. These benefits are real. So are the costs of restricting training data.
But there are also costs to the status quo that this post doesn't grapple with. If AI systems can be trained on copyrighted work without permission or compensation, and those systems then compete with the original creators in various ways, that shifts the economics of creative work. Maybe that's a shift we should accept in the name of progress. Maybe it isn't. But pretending there's no trade-off is intellectually dishonest.
The Chamber of Progress brief, excerpted in the post, talks about how "musicians use generative AI to explore arrangements, develop melodies, and bring projects within reach that limited resources might otherwise foreclose." Great! Also true: some musicians are competing with AI-generated music trained on their work without compensation. Both things can be true.
Where This Goes
We won't know how the court rules for months, possibly years if there are appeals. In the meantime, expect more moves like this—both sides trying to shape the narrative, marshal expert support, and frame the stakes in the most favorable light.
What's worth watching: whether other tech companies facing similar lawsuits adopt OpenAI's playbook of aggressively public counter-messaging, and whether creators develop equally sophisticated advocacy strategies in response.
The legal question is whether AI training is fair use. The meta question is who gets to define what's at stake—innovation and accessibility, or the sustainability of creative work. OpenAI's blog post is a bet that they can win the second fight even while the first one is still undecided.