Sign up for The Media Today, CJR’s daily newsletter.
Last week, leading figures in the media industry met at Columbia Journalism School for a two-day event, hosted by the Tow Center for Digital Journalism and the Craig Newmark Center for Journalism Ethics and Security, focused on the role of generative AI in the newsroom. Technology reporters from publications including the New York Times, 404 Media, and The City were present, along with academics and AI experts. As part of the event, the Tow Center revealed preliminary findings from a forthcoming report on the relationship between news publishers and platforms in the AI era, drawing on more than thirty interviews with newsroom leaders, platform executives, and academic experts. (A livestream of the event is still available on YouTube.)
One of the main points of discussion throughout the event was the (increasingly strained) relationship between news publishers and tech platforms. For decades, the former have been trying to get the latter—in particular, Google and Meta—to compensate them for the news they distribute and profit from. Progress was made, at least on the news side, in 2021, when Australia introduced a first-of-its-kind “tech tax,” a law that resulted in around two hundred and fifty million Australian dollars flowing into newsrooms yearly. According to Anya Schiffrin—the director of the technology and media specialization at Columbia University’s School of International and Public Affairs, and a panelist at the event—this success made Google and Meta worry that the laws would spread. In a raw exercise of corporate power, Schiffrin alleged, tech platforms started “throwing money and lawyers” at similar bills elsewhere, to get them killed.
From there, the situation became even more complicated: when Canada followed suit with the Online News Act, Meta refused to play ball, pulling its funding for journalism in Canada and blocking news on Facebook and Instagram. (Lauren Watson wrote about the effects of the Facebook news ban for CJR this week; we’ll have more on that in tomorrow’s newsletter. In the end, Google agreed to pay Canadian publishers one hundred million Canadian dollars annually to link to their articles.) In the US, California considered adopting a similar law to reverse the decline in local media, as Cameron Joseph wrote for CJR earlier this year. But in July, lawmakers in the state decided to shelve the legislation and instead commit two hundred and fifty million dollars over five years to bolster local journalism, with Google covering the majority of the sum. An inadequate figure, according to Schiffrin: “It went from something that should have been billions of dollars into absolute peanuts.”
As we have entered the era of generative AI, the relationship between platforms and publishers has continued to sour. Perhaps most notably, the New York Times sued OpenAI, arguing that it infringed the paper’s copyright when it trained its large language models on millions of Times articles without permission. In an apparent attempt to appease publishers, AI companies are now writing checks to newsrooms to license their content for training purposes. As my colleague and Tow research director Pete Brown has written, the big question for many newsrooms becomes: “Deal or no deal? (For those not invited to the table, it’s more a case of: Deal with it.)”
Whether to take such deals—and if they’re even suitable for the news industry—was another key talking point at the event. “If we’ve learned anything in this business in the last ten years, it’s that we have to be aggressive about owning our own technology and staying up with what’s changing. But that costs money,” Kimberly Lau, vice president of consumer media and president of Scientific American, said. “I wouldn’t say that I absolutely would do a deal tomorrow, but I would much rather be engaging and having a conversation and have a seat at the table than be left out of the conversations altogether.” (Currently, Scientific American is not party to any licensing deal.)
With newsrooms already in a rough spot after years of sweeping job losses, licensing and revenue-sharing deals can feel like a tempting solution, at least in the short term. And despite some initial hesitation, news publishers are increasingly deciding to sign such deals: “News Corp, the Financial Times and Dotdash Meredith became the latest news organizations to strike licensing deals with OpenAI, while Axel Springer and Informa teamed up with Microsoft,” Brown wrote in May, as part of an effort to track these partnerships. Since then, the AI startup Perplexity has announced revenue-sharing deals with major news outlets, including Time, Der Spiegel, and Fortune. According to The Information, OpenAI has offered some media firms as little as between one and five million dollars a year to license their news articles for training purposes. (Some of the deals have included other, in-kind benefits.)
The event also saw a discussion of research conducted by Brown and Klaudia Jaźwińska, on the evolving relationship between publishers and platforms in the AI era. Their findings, which will be published in a forthcoming report, draw on over thirty interviews with newsroom leaders, platform executives, and academic experts on their views of licensing deals and AI experimentation. Regardless of the topic, Brown said, most conversations returned to the issue of intellectual property. “This question remains sort of unresolved even as tech companies are racing to develop these search products that use AI models to summarize news,” he said.
Publishers who spoke with Tow said that there is a need for regulatory intervention, ideally as soon as possible. But Big Tech firms, in their efforts to dominate the AI sector, would prefer to avoid anything that slows them down. According to Tow’s research, this creates a misalignment over speed: publishers want to establish copyright rules and regulations, while tech platforms are reluctant to move at all. “Big Tech only moves with a gun to its head,” one former platform executive bluntly told Tow.
The second recurring theme revolves around geographical issues: even when platforms do try to accommodate news publishers, there are very different requirements depending on the country in which they operate. The complexity of this arrangement, according to one former platform executive, drove senior leadership “nuts.” Finding common ground between platforms and publishers may ultimately require simple, specific solutions that address such complexity on the platform side while acknowledging the journalistic value presented by newsrooms, according to Tow’s research.
The rapid development of generative AI hasn’t just (further) strained the relationship between platforms and publishers; it has also reshaped the role of journalists—a topic of discussion during the event’s final panel. Ben Smith, the cofounder and editor in chief of Semafor, said that one of the core threats of AI is the “Liar’s Dividend”—a term that refers to politicians falsely claiming that their unethical behavior is misinformation or fake news. “It’s not so much that we’re being tricked by the pope in Balenciaga, but that when someone denies something real, they can get away with it,” Smith said (referring to a convincing, yet fake, image of the pope in a puffer jacket that went viral last year). The biggest impact of AI on journalism is having to verify or debunk a tide of inaccurate content, Julia Angwin, the founder of Proof News and a New York Times contributing opinion writer, said. “There’s a hope for journalism, but it’s also a sad hope: that we’re just low-paid fact-checkers for Big Tech.”
Other notable stories:
- Yesterday, the judge in Donald Trump’s federal election-subversion case published a detailed filing by prosecutors that aimed to present the charges against Trump as compatible with a recent Supreme Court ruling granting presidents broad presumptive immunity for official acts, but also offered new evidence about Trump’s behavior on and around January 6, including his reply to an aide who told him that his vice president, Mike Pence, was in danger during the insurrection: “so what?” “In recent American history, there are probably few moments that have been more scrutinized,” Politico notes this morning—and yet Smith was able to go further than other investigators. The filing was partially redacted, but some details were decipherable by cross-referencing it with an old Trump tweet (which the Post’s Michael Scherer dubbed “the Rosetta tweet”).
- This week, Katherine Jacobsen, of the Committee to Protect Journalists, published a report assessing the state of press freedom in the US and how the election could affect it. “Interviews with journalists, lawyers, and press freedom advocates over several months ahead of the November vote found that media workers are confronting challenges that include an increased risk of violence, arrest, on- and offline harassment, legal battles, and criminalization,” she reports. (One described January 6 as “a warning shot.”) “Overseas, journalists fear that a second Trump term would again embolden foreign leaders to restrict their own media, negatively affecting the global press freedom landscape and undermining those in regions that rely on U.S. aid and support.”
- For Reason, C.J. Ciaramella reports on a recent court ruling that banned the Ocala Gazette, a newspaper in Florida, from publishing surveillance footage of a 2022 incident in which a man who was mentally ill “died in the Marion County Jail after deputies pepper-sprayed, dogpiled, and tased him.” The Gazette, which has now been able to view the footage after a lengthy battle, reports that the man exhibited “no physical violence” toward the jailhouse officials (the local sheriff’s office disputes this), but a judge ruled that putting the footage itself online would “raise safety concerns,” Ciaramella reports. The episode, he adds, “illustrates the importance of local news outlets.”
- In this newsletter last week, we wrote about the case of Pavel Durov, the CEO of Telegram who was recently arrested in France on charges related to criminal activity on the app. Now the New York Times has an interview with Irina Bolgar, a longtime former romantic partner of Durov’s who has alleged in a separate criminal complaint in Switzerland that he was abusive toward their son. (Durov has strongly denied this.) Bolgar’s testimony “opens a window into the rarefied life” of Durov, “who remains one of the world’s least understood tech magnates,” Adam Satariano and Paul Mozur report.
- And Le Monde reports on the stalled progress of a center dedicated to press cartooning in Paris, which was first proposed by Georges Wolinski—a caricaturist who worked at Charlie Hebdo until he was killed in the terrorist attack on its offices in 2015—and has since been supported by senior officials, including President Emmanuel Macron. Now, however, the government has gone quiet on the plans. According to Le Monde, one possible reason is concern among security officials of another terrorist attack.
ICYMI: Confronting falsehoods carries risks for the press. So does ignoring them.
Sarah Grevy Gotfredsen is a computational investigative fellow at the Tow Center for Digital Journalism at Columbia University. She works on a range of computational projects on the digital media landscape, including influence operations conducted through news media and the information ecosystem. She graduated from Columbia University in 2022 with an MS degree in data journalism.