Insurance against the courts, dressed up as partnership
In July 2023, OpenAI and The Associated Press formalized an agreement granting the AI company licensed access to nearly four decades of journalism in exchange for technology expertise — a quiet transaction that speaks to a larger reckoning underway. As machines grow capable of producing language indistinguishable from human writing, the question of whose words trained them has become urgent, legal, and deeply moral. This deal represents one answer: not seizure, but negotiation — a tentative model for how knowledge economies might adapt when the machines come to the table.
- AI companies face mounting legal pressure as thousands of authors and news organizations accuse them of harvesting copyrighted material without consent or compensation.
- The FTC opened an investigation into OpenAI on the very day the AP deal was announced, signaling that regulatory scrutiny is no longer hypothetical.
- OpenAI's licensing agreement with AP functions as strategic insurance — a way to demonstrate legitimacy and secure training data even if courts restrict access to freely scraped content.
- AP gains not money but influence: a seat at the table with one of the most powerful AI companies, and the chance to shape how generative tools enter the newsroom.
- The deal sets a precedent that could ripple across the news industry, offering thousands of AP-affiliated outlets a template for monetizing archives rather than watching them be consumed for free.
In July 2023, OpenAI and The Associated Press announced a licensing agreement giving OpenAI access to AP's news archive dating back to 1985 — decades of reporting, sports recaps, and breaking news. In return, AP received access to OpenAI's technology expertise and an invitation to explore what generative AI might offer journalism. Financial terms were not disclosed.
The deal arrived amid genuine turbulence. ChatGPT had upended expectations of what machines could do with language, but those machines required enormous quantities of written material to learn from. OpenAI and its peers had been scraping the internet for years, and the consequences were catching up with them. More than four thousand authors — including Nora Roberts, Margaret Atwood, and Jodi Picoult — signed an open letter accusing AI companies of taking their words without payment. Others, like Sarah Silverman, filed lawsuits.
News organizations occupied an uneasy position. They had been exploited by tech platforms before, and yet many saw potential in AI tools. AP itself had spent years using automation to write earnings reports and sports summaries. When OpenAI came calling, there was reason to engage.
For OpenAI, the deal was a hedge. Northwestern professor Nick Diakopoulos observed that the company was insulating itself against the courts — securing legitimate training data it could point to if lawsuits succeeded in restricting access to scraped material. For AP, the calculus was more exploratory: no generative AI in the newsroom yet, but curiosity about what might come.
The broader significance lay in precedent. AP's content reaches thousands of news outlets worldwide. If it was willing to license its archive, others might follow — establishing negotiated agreements as the industry norm rather than litigation or silent extraction. Whether the compensation offered would prove adequate remained unresolved.
The timing sharpened the stakes. On the same day the deal was announced, the FTC opened an investigation into OpenAI over data scraping and potential misinformation. The regulatory machinery was in motion. OpenAI's agreement with AP read, in that light, like a company trying to demonstrate it could play by the rules — before the rules were written for it.
On a July afternoon in 2023, OpenAI and The Associated Press announced they had struck a deal. OpenAI would gain access to AP's archive of news stories stretching back to 1985—decades of reporting, earnings announcements, sports recaps, and breaking news. In exchange, AP would receive something less tangible: access to OpenAI's technology expertise and a seat at the table as the company explored what generative AI might do for news products. The financial terms stayed private.
The arrangement arrived at a moment of genuine turbulence for AI companies. ChatGPT, released the year before, had upended the landscape—suddenly there were machines that could write, summarize, and synthesize text in ways that felt almost human. But those machines needed to eat. They required vast quantities of written material to learn from: books, articles, social media posts, anything with language in it. OpenAI and its peers had been scraping the internet for years, and now the bill was coming due.
Authors were the first to organize. Over four thousand writers—Nora Roberts, Margaret Atwood, Louise Erdrich, Jodi Picoult among them—signed a letter to the CEOs of OpenAI, Google, Microsoft, Meta, and other AI makers. They accused these companies of theft dressed up as innovation: taking their words, their style, their ideas, and feeding them into machines that could now produce text that sounded like them, without paying them a cent. Some authors went further. Sarah Silverman and others filed lawsuits. The legal machinery was grinding to life.
News organizations occupied an awkward middle ground. They had been burned before by tech companies—Facebook had promised them traffic and then changed its algorithm; Google had built a search empire partly on their reporting. But they also saw something in AI that might be useful. AP itself had been experimenting with automation for nearly a decade, using algorithms to write earnings reports and sports summaries. The organization had even built an AI-powered image search tool. So when OpenAI came calling, there was reason to listen.
The deal made strategic sense for both sides, though in different ways. For OpenAI, it was insurance. Nick Diakopoulos, a professor at Northwestern who studies the intersection of communications and computer science, put it plainly: the company was hedging its bets against the courts. If lawsuits succeeded in cutting off access to freely scraped material, OpenAI would still have a guaranteed supply of legitimate training data. It was a way of saying: we have permission. We paid for this. We are doing this right.
For AP, the calculus was more complex. The organization wasn't using generative AI to write its news yet—that remained controversial, and for good reason. But it was curious. The statement mentioned exploring "potential use cases for generative AI in news products and services," though it offered no specifics. Both companies pledged to pursue "responsible creation and use" of these systems, language that had become almost mandatory in any AI announcement by mid-2023.
What made this deal significant extended beyond the two parties involved. Ken Doctor, a news industry analyst, noted that AP's reach was enormous. It supplied stories to thousands of newspapers and broadcasters. If AP was willing to license its archive to an AI company, other news organizations might follow. The precedent mattered. It suggested a path forward—not free access, not lawsuits, but negotiated licensing deals where news organizations got something in return. Whether that something was enough remained an open question.
The timing was also notable. On the very day OpenAI and AP announced their agreement, the Federal Trade Commission opened an investigation into OpenAI. The agency wanted to know whether the company had violated consumer protection laws by scraping public data without permission and whether ChatGPT had published false information. The FTC declined to comment. But the message was clear: the regulatory machinery was grinding to life too. OpenAI's deal with AP looked like a company trying to get ahead of the storm.
Citações Notáveis
In order to guard against how the courts may decide, maybe you want to go out and sign licensing deals so you're guaranteed legal access to the material you'll need— Nick Diakopoulos, Northwestern University professor
The deal's effects could reach far beyond the AP because of the organisation's size and its deep ties to other news outlets— Ken Doctor, news industry analyst
A Conversa do Hearth Outra perspectiva sobre a história
Why does OpenAI need AP's archive specifically? Isn't the internet full of news?
The internet is full of news, yes, but it's also full of lawsuits now. OpenAI can say it scraped freely, but courts might disagree. A licensed deal is insurance—it's proof of permission.
And what does AP get out of it?
Technology expertise, they say. But really, they get a seat at the table while AI reshapes their industry. They're not fighting it; they're trying to shape it.
Is this good for journalism?
That depends on what AP does with it. If they use it to automate away reporting, no. If they use it to help reporters work faster, maybe. The deal itself doesn't answer that question.
What about the authors who are suing?
They're fighting a different battle. They want compensation for their words being used to train machines. AP is trying to get ahead of that by negotiating. Authors are saying: you should have asked first.
Will other news organizations follow?
Almost certainly. If AP can get something for its archive, why wouldn't The New York Times or The Washington Post try? This deal sets a precedent.
What's the real story here?
It's about who controls the transition to AI. Companies like OpenAI want to move fast. News organizations and authors want to be compensated. This deal is one attempt to bridge that gap—but it's just the beginning.