Your work has value, and using it requires permission and payment
In mid-July, OpenAI and The Associated Press formalized a licensing agreement that places a quiet but consequential stake in the ground: the words that teach machines to speak belong to someone, and that someone deserves acknowledgment. The deal grants OpenAI access to AP's storied news archive while returning to AP both compensation and a foothold in emerging AI technology. It arrives amid a widening cultural reckoning over whether the vast human record of thought and expression can be harvested freely in service of artificial intelligence — and who bears the cost when it is not.
- AI companies have been training powerful language models on oceans of human-written text, often without permission or payment, and the people whose words were used are beginning to demand accountability.
- More than four thousand authors — including Margaret Atwood and Nora Roberts — signed an open letter accusing AI firms of exploitation, while lawsuits from novelists and Sarah Silverman have escalated the conflict into courtrooms.
- The OpenAI-AP licensing deal attempts to chart a different course: negotiated consent and mutual benefit rather than silent extraction and legal battle.
- AP frames the agreement as a potential industry model, signaling that intellectual property protections and fair compensation for creators could become the expected standard — not the exception.
- Whether this deal inspires a wave of similar agreements or remains an isolated arrangement between two well-positioned institutions is the question now hanging over the entire AI and media landscape.
On a Thursday in mid-July, OpenAI and The Associated Press announced a licensing agreement that cuts to the heart of one of artificial intelligence's most contested questions: who owns the words a machine learns from?
Under the arrangement, OpenAI gains rights to portions of AP's extensive news archive, while AP receives access to OpenAI's technology and expertise. Financial terms were not disclosed. The deal arrives as large language models like ChatGPT continue to require vast quantities of text to function — and as the companies building them face growing scrutiny for harvesting that text without permission or payment.
The backlash has been vocal. Over four thousand writers, including Nora Roberts, Margaret Atwood, and Jodi Picoult, signed a letter last month accusing AI companies of exploitative practices. Novelists and comedian Sarah Silverman have filed copyright infringement lawsuits against OpenAI. News organizations have raised parallel objections, arguing their reporting should not fuel commercial AI systems for free.
AP's senior vice president Kristin Heitmann framed the agreement as more than a transaction — as a statement of principle. Her language positioned the deal as a potential model for how the industry should treat creators: with recognition, respect, and fair compensation.
Both organizations said they intend to explore generative AI applications in news products, though no specifics were offered. AP itself has used narrower forms of AI for nearly a decade, automating earnings reports and sports recaps, but has not yet introduced generative AI into its journalism.
The deeper question remains open: will this agreement become a template that reshapes how AI companies compensate the human minds whose work trained them — or will it stand as a rare exception in an industry still largely resistant to that reckoning?
On a Thursday in mid-July, OpenAI and The Associated Press announced they had reached a licensing agreement—a move that signals how the artificial intelligence industry is beginning to reckon with the question of who owns the words it learns from. Under the deal, OpenAI gains the right to license portions of AP's vast archive of news stories. In return, AP gets access to OpenAI's technology and product expertise. Neither organization disclosed what OpenAI is paying for this arrangement.
The agreement arrives at a moment of genuine friction. Large language models like ChatGPT require enormous quantities of text to function—books, news articles, social media posts, anything with words. The companies building these systems have been vacuuming up written work from across the internet, often without explicit permission or compensation. Last year's release of ChatGPT sparked a wave of generative AI products that can write new text, generate images, and create other media. The technology is powerful and increasingly useful. It is also prone to confidently stating things that are false, and it raises a straightforward question: if a machine learned to write by studying someone else's work, who should be paid?
Authors have begun pushing back. More than four thousand writers—including Nora Roberts, Margaret Atwood, Louise Erdrich, and Jodi Picoult—signed a letter last month to the leaders of OpenAI, Google, Microsoft, Meta, and other AI companies, accusing them of exploitative practices. The letter described chatbots that "mimic and regurgitate" the language, style, and ideas of human writers. Some novelists and the comedian Sarah Silverman have filed lawsuits against OpenAI for copyright infringement. News organizations have raised similar objections, arguing that their reporting should not be used to train commercial systems without compensation.
The OpenAI-AP deal appears to be an attempt to move past this impasse through negotiation rather than litigation. Kristin Heitmann, AP's senior vice president and chief revenue officer, said in a statement that the organization is "pleased that OpenAI recognizes that fact-based, nonpartisan news content is essential to this evolving technology, and that they respect the value of our intellectual property." She added that AP "firmly supports a framework that will ensure intellectual property is protected and content creators are fairly compensated for their work." The language suggests AP views this agreement as a model—a way of establishing that creators deserve payment when their work trains AI systems.
Both companies also said they are exploring how generative AI might be used in news products and services, though they offered no concrete examples. OpenAI and AP stated they "believe in the responsible creation and use of these AI systems." The AP itself has not yet deployed generative AI in its news stories, but it has used other forms of artificial intelligence for nearly a decade—automating corporate earnings reports, recapping sports events, and powering an image archive search tool. The organization also runs a program helping local news outlets incorporate AI into their operations.
What remains unclear is whether this deal will become a template for other negotiations, or whether it will be seen as an outlier—a special arrangement between two organizations with particular leverage and alignment. The broader question persists: as AI becomes more central to how information is created and distributed, how will the people and institutions whose work trained these systems be compensated? The OpenAI-AP agreement suggests at least one answer is possible. Whether it becomes the standard is another matter entirely.
Citas Notables
AP firmly supports a framework that will ensure intellectual property is protected and content creators are fairly compensated for their work.— Kristin Heitmann, AP senior vice president and chief revenue officer
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that OpenAI and AP made this deal public? Why announce it at all?
Because it breaks a pattern. For months, AI companies have been training on text without asking permission or paying. This deal says: we negotiated. We paid. It signals that the old model—just take what you need—might not hold anymore.
But the price is secret. How do we know AP got a fair deal?
We don't. That's the real tension. AP is claiming victory, but without knowing the number, it's hard to judge. It could be substantial or symbolic. The secrecy itself is telling—both sides wanted to avoid setting a precedent they'd have to defend.
What about all those authors who sued? Does this help them?
Not directly. But it establishes a principle: your work has value, and using it requires permission and payment. That's what the lawsuits are arguing for. If this deal holds up and becomes common, the authors win. If it stays an exception, they're still fighting.
Will other news organizations demand the same deal?
Almost certainly. Once AP gets paid, every other news outlet will ask why they shouldn't. OpenAI will face the same question from book publishers, music labels, everyone whose content trained their models. This deal might look like a solution, but it could also be the beginning of a much larger reckoning.
What's the real story here—the deal itself, or what it reveals?
What it reveals. The deal is a negotiation between two organizations with power. The real story is that AI companies can no longer simply take. They have to ask. That's the shift.