When the nation’s largest newspaper chain announced a deal with Palantir in August, USA Today Co. CEO Mike Reed framed it as a necessary modernization step for a print-heavy publisher. The company, which operates USA Today and more than 200 local publications, still generates $1.1 billion in annual print revenue, but faces the same circulation declines and changing digital habits battering the rest of the industry.
Reed laid out the AI-fueled vision at a Palantir conference last month. The company wants to consolidate years of fragmented reader data into a single system and use AI to personalize newsletters, target affiliate links, improve content discovery, and turn anonymous visitors into registered subscribers. Demoing the platform with an “anonymized” user named “Marcus,” Reed described a continuous feedback loop built around reader behavior. “We’re actually an intelligence organization now,” he said, “because of the shift to online and the data we have on consumers.”
Last month, CEO Mike Reed demoed Palantir’s platform analyzing an example “anonymized” reader [Photo:
Palantir]
The reaction from the company’s journalists was swift and, perhaps, unsurprising. The deal, announced during an August investor call and first reported by Nieman Lab, prompted a revolt across 31 unionized newsrooms represented by the NewsGuild-CWA. In a joint letter calling for the contract to be terminated, staffers argued that Palantir’s work with U.S. Immigration and Customs Enforcement creates a conflict for reporters covering immigration enforcement and protests, while raising questions about reader data and public trust. In addition to a booming commercial business, Palantir—fronted by its chest-thumping CEO Alex Karp—also holds billions of dollars in federal contracts, including for the Defense Department’s flagship data fusion platform. “Every problem we have with this partnership really comes down to trust,” Mike Davis, an investigative reporter at the Asbury Park Press, told Straight Arrow News.
That skepticism comes after years of cuts. Since the company formerly known as Gannett merged with GateHouse Media in 2019, it has gone through repeated rounds of layoffs, essentially halving its workforce by 2024, even as CEO compensation rose amid flat or declining median employee pay. The company had roughly 7,500 U.S. employees last year, about 15% of them unionized, according to its latest SEC filing. Leadership has cast the restructuring as a response to falling readership and new forms of distribution and discovery, including AI feeds and chatbots. Much the same logic now underpins its AI push and its alliance with one of the world’s most controversial technology companies.
USA Today isn’t the only media organization to partner with Palantir, or face scrutiny over surveillance. Thomson Reuters has long supplied data used in federal immigration surveillance tools. Marketing firms Stagwell and Zeta Global have worked with Palantir on ad targeting. Fox News Digital used Palantir to build an internal platform that supports much of its article production, while Axel Springer, owner of Politico and Business Insider and many other outlets, has partnered with the company to better understand audience behavior.
Last month, Axel Springer also gave Palantir’s cofounder and chairman Peter Thiel its 2026 Axel Springer Award, recognizing him “for his transformative impact on the digital age.” A decade earlier, Thiel, who also founded PayPal, secretly bankrolled Hulk Hogan’s lawsuit against Gawker Media, contributing to the publication’s bankruptcy and demise. Some USA Today staffers see that episode as part of a broader pattern of Thiel-linked hostility toward the press, and another reason to ditch the contract.
Reporters have also lately raised concerns about the impact of AI on their work and their livelihoods. USA Today-owned newsrooms are using AI tools for research and investigative tasks, like filing public records requests, and to generate headlines and automate parts of digital publishing. After an abortive experiment in AI-written sports stories starting in 2023, the company the following year rolled out AI-generated “key points” that sit at the top of stories. USA Today Co. policy says that if any “additional AI-generated content is published, journalists must include AI disclosure language for transparency and “must be verified by a human for accuracy prior to publication.” Last year, WGBH reported that several of the company’s Boston-area news publications were using an AI tool called Espresso that is “designed to draft polished articles from community announcements.”
At the same time, the company is feeding its journalism into generative AI products, including its DeeperDive answer engine and through distribution and licensing deals with Meta, Amazon, and Microsoft. (It has also sued Google, along with the Daily Mail and a class of about 5,000 other publishers who allege the search giant overcharged them through its advertising technology platform, and on Thursday, it sued OpenAI for allegedly stealing its content to train its large language models. ) As Reed put it on the investor call: “We recognize that we have to create and format content for humans and for machines.”
In an interview last month, Reed, Media President Kristin Roberts, and communications executive Lark-Marie Anton spoke about the rationale for the Palantir deal, how AI is actually being used in the newsroom, and the union friction it’s provoked. Contrary to what they called falsehoods by union leaders, they said Palantir’s software, assembled with the help of a team of embedded Palantir engineers, works only with governed, “anonymized” data, and said the company has no direct access to newsroom systems, reporter notes, or subscriber records. Instead of threatening journalists’ jobs, Reed said AI tools could make the newsroom and business more productive and profitable, and, he hoped, would allow the company to hire more journalists.
Following the publication of this interview, NewsGuild President Jon Schleuss responded in a statement: “USA Today Co. executives’ animosity toward their workers and unions is nothing new. The NewsGuild, which represents over 800 unionized USA Today Co. workers, fully stands behind the statement workers released when this deal with Palantir was first announced.”
This conversation has been lightly edited for clarity. Anton provided some additional written answers shortly before she departed the company on October 1.
How did USA Today Co. come to work with Palantir? What was the spark for this relationship?
Mike Reed: I think that just being connoisseurs of the media, we knew some of the problems we had with having a lot of data, but also having multiple systems and not being able to have our data cohesive, in one place, and being able to learn from it and have recommendations from it. And so understanding that problem and talking about it internally and thinking, well, this is going to take years and it’s going to be expensive to maybe get ourselves to a place where we can use this in a real business-oriented way. And watching a lot of interviews with CEOs who were partners with Palantir saying that with Palantir, they did in a matter of a month or two things that might have taken them years to do. And I just thought, we need to talk to them, because whether it’s cheap, expensive, whatever it is, we have a business issue to contend with, and trying to build something ourselves that might take three or four years—and we don’t even know if we’re going to get it right and it might cost a lot of money. Or, we could partner with somebody like Palantir, and be ready next month.
And so that’s why we said let’s talk to them. And we liked what we heard from them, because it was obvious that they were going to be execution-oriented, not just, we’re going to think about your problem and tell you what to do. They’re like, we’re going to actually do the work and help you execute. And that was really refreshing. So we said let’s do it.
Did you consider other companies?
Anton: We evaluated the landscape broadly because our challenge is not simply analytics, personalization, or AI. It’s the need to connect a very large, complex set of audience, advertising, subscription, commerce, and distribution data into a common operating environment. What stood out about Palantir was its ability to create what it calls an ‘ontology’ layer—essentially a unified model that allows different datasets and systems to work together and be acted upon in real time. That capability was particularly relevant to a company like ours because we already have significant first-party data and strong audience reach. The challenge isn’t collecting more data, but making better use of the data we already have. As Mike has said internally, Palantir is operationally more sophisticated in this area than any vendor we reviewed.
CEO Mike Reed [Photo: Courtesy USA Today]
Is there a metric you’re focused on here, or is there something that you think so far has proven to be valuable?
Reed: Not yet is the short answer. We started working with Palantir in June. So we really only have a couple months under our belt, and most of that time has been them really organizing our data, trying to take all this data from disparate systems and make sense of it in a manner where they can use it, they can provide the business outcomes we’re looking for. When we started to engage with them, we came up with a list of, here’s the business outcomes we’d like from you. Here are the problems we’re solving for. And then they worked with our data to try to figure out exactly what the solutions would be. So that’s really where the first couple months have been. But what might have taken us a couple of years, they figured out shortly, and will be in the market here as we go into the fourth quarter and start to, hopefully, have some positive learnings.
What are some of those outcomes you’re aiming for?
Reed: One is, how do we engage longer with consumers on our platform? We don’t want consumers to come for one story and leave. So how do we make the experience more enjoyable and more personal? Our audience is massive. A lot of times they’re there for a couple minutes and then they’re gone. And so how do we keep them there, how do we engage with them and how do we get them back more often?
And the second thing we’re trying to solve for is, we have a lot of users that come to our platform whom we don’t know anything about. They’re not registered, they’re not subscribers. And so while we can still derive some value from them, it’s beneficial. If we can create known users out of unknown users, the value we can create from our relationship with that consumer goes up exponentially.
What is Palantir being paid, and how does that compare to what an in-house build would have cost? What’s the length and exclusivity of the contract, and is there an off-ramp if the results you’re after don’t materialize?
Anton: I will share that we approached this with the same financial discipline we apply to any strategic technology investment. We evaluated the expected return, the speed to value, the opportunity cost of building internally, and the resources required to build, support, and maintain a comparable platform ourselves. And we have been focused from the outset on flexibility and long-term independence. One of the key workstreams has been ensuring knowledge transfer and building internal capability so that we are not dependent on outside engineers forever. We have also evaluated off-ramp scenarios and continuity plans as part of the process: If we don’t see measurable business value, we will reassess just as we would with any strategic investment.
How do you think about USA Today’s goals as part of a larger structural shift that’s happening across news media—or a shift that you think will need to happen?
Reed: I think a big structural shift for news media like ours is that, when we think about content creation, it can’t just be hard news. We have to create a lot of sports content and entertainment and games and puzzles. If we’re going to engage with consumers for longer periods of time and more often, we have to have a broader content strategy than just news. So that’s a big structural change for an industry that’s really created news historically, and not really focused on a lot of that other stuff. And that other stuff’s really important now.
And the second thing is listening to the data, in terms of your content creation strategy. And if you think about it, not that long ago, your newsroom would meet in the morning around a table and say, ‘What are we going to cover today?’ And you would go cover it, and that’s the news that would be reported. And our people sitting around the newsroom, they don’t have bad ideas, but why not listen to the consumers? What do they want covered? And so looking at the data and understanding what’s resonating with consumers, what are they focusing on and doing more of that and looking at the data and saying, ‘They’re not paying any attention to this, maybe we should do less of that or none of that.’ That will allow us, I think, to create more content that’s engaging to consumers.
Don’t mistake that to mean that we would abandon the social mission side of our business, which is to cover the communities we operate in, cover the country, do investigative work and do the things that are going to protect citizens and people who live in our communities. Of course we’re going to do that work, but there’s hard news that we cover that’s not necessarily aligned with that mission that people don’t care about. So let’s do less of it.
How are people in the newsrooms responding to these changes? What questions do they have, and how are you responding to them?
Reed: I think that a lot of folks in our newsroom are covered by the NewsGuild, and I think the NewsGuild gives them a lot of bad information that leads them to decision-making that is not the best. And so I think the News Guild is doing a disservice to the people, its own members who are our employees and that are part of this cohort of journalists that have to make these changes. And I think that makes it harder, and I think we’re having more success in newsrooms that are not unionized, with this new world, than those that are unionized.
Kristin Roberts, USA Today’s Media President [Credit: USA Today]
Kristin Roberts: Guild leadership is saying things that are factually untrue. They prey on fear. No matter that I will appear in front of any newsroom for a standup in which they ask me anything for an hour and a half at a time, and I explain to them all of the things. Every single journalist in the organization has my cell phone number and they can call and say, I just heard this from the Guild rep. Is this true? And I’ll tell them it isn’t. I’ll give you an example that’s incredibly real for us. Palantir doesn’t work inside of any USA Today Company system. They don’t have access to any of the notes our reporters take. They’re not in the Microsoft suite of tools. They don’t have information directly about our consumers or our staff. I said this to all of the staff, took their questions, and answered this precisely. They don’t work in our systems. We feed them highly controlled data. We’re in their shit. They’re not in ours.
Reed: Two other untruths before we move on. There’s this fear—and I think the Guild plants this as well as outside cynics—is that we’re going to replace journalists with AI. That couldn’t be further from the truth. We’re hoping that our journalists can do more of the work they really want to do. And as we’re successful with our business strategy, we’re going to be able to hire more journalists. So this is not an exercise to replace journalists with AI at all. It’s hopefully going to allow us to add more journalists. Second is that I think the union plants fear that we’re going to have AI-created content, even though we say till we’re blue in the face we’re not.
Reporters have also raised concerns about data protection. How do you ensure the data is protected when working with Palantir or any vendor? Are there independently verifiable guardrails that prevent sensitive data from being accessible to Palantir’s broader data environment?
Anton: This was one of the most important areas of diligence. Our position is straightforward: our data remains our data. We maintain ownership and control over it, and we require vendors to meet stringent security, privacy, and governance standards. Protecting personal information is fundamental to maintaining audience trust, and those requirements apply to Palantir just as they apply to any strategic technology partner. Data governance, access controls, ownership, security review, and legal protections were central considerations in negotiating the agreement and our ongoing working relationship.
Palantir burnished its reputation for years by staying clandestine, and some of its work is pretty controversial. How do you think folks, especially readers but also reporters and their sources, should think about this partnership in light of its shadowy reputation? Was there discussion about the reputational risks involved here?
Anton: It’s a fair question. We recognize that Palantir has a public profile that prompts strong opinions. That’s part of the reason we’ve been deliberate about discussing not only the opportunities but also the governance, privacy, and trust considerations involved. For us, the key question wasn’t who else Palantir works with. The question was whether the partnership advances our mission while meeting our standards. Our journalism remains independent, and our editorial decisions are governed by our longstanding ethics and standards policies, and those policies do not change because we adopt a technology platform.
We also work with other technology companies that serve governments, enterprises, and other complex organizations. We evaluate partners based on their capabilities, the strength of their controls, and whether the relationship aligns with our values and business objectives. Ultimately, readers should judge us by our journalism, our transparency, and how responsibly we steward their trust.
Are there other particular ways that you see AI helping inside the newsroom? And is that something that Palantir has been working on too?
Roberts: A lot of the things that we have as reporters, over the course of my 30 years in journalism—I started as a news assistant at Reuters on the Wall Street desk—I have been asked to do more and more and more tech stuff, applying tags, applying quotes, inserting links, putting ads in different places. None of that has to happen anymore. Think of all the times with your editor where you were stumped on the right headline and you came up with 17, and you had to A/B test them. Now the system will give you A/B testing in real time. All of this stuff actually frees our mind to be creative, because you’re no longer spending 27 minutes searching for a lead video, wasting your time.
I’ll give you another example. In all of our entertainment stories and our sports stories, we do affiliate links to our partners. So if you’re reading a sports story, you want to buy the jersey for that team, our reporters and editors used to have to go to a file, find a relevant link, copy that link, put it in. All of that is now automated, and that’s a Palantir project.
Can Palantir’s software help with social media and promotion? Can it help with ad targeting? Can it help reporters directly with their reporting, or could it eventually?
Anton: I think the potential applications extend well beyond those initial examples. On the audience side, the goal is to better understand what a reader values and surface the most relevant experience at the right moment. For advertising, it can help connect audience behavior, content engagement, advertiser demand, and first-party data to improve targeting, relevance, inventory valuation, and monetization decisions. Hypothetically, could it support social strategy? Potentially. Any system that helps us better understand audience interests and content performance can inform distribution decisions across channels.
Again, hypothetically, could it help reporters directly? Perhaps over time. Other media organizations have used Palantir-powered tools to assist journalists with things like briefings, research support, and workflow tools. But the task at hand today is focusing on the audience and consumers to drive revenue and our business transformation.
How are your reporters using generative AI now?
Roberts: I think the vast majority of people in the company, and certainly in the newsrooms, are the ones adopting AI enthusiastically. So we’ve pushed all of our AI tools directly to our staff and say, Have some fun, figure this out. How can this make your job easier and better so that you can get back to the thing that you and I got into journalism to do, which is to break news and deliver scoops? All of our problems in journalism are solved if we get back to exclusive and distinctive, because AI answer engines can’t do that.
For example, one of our news assistants in the UK created FOIAbot, and we have now been able to deploy that across the entire company. Say, I work in Naples, Florida, and I’m a local reporter. I want to see how many accidents happen at the intersection of First and Main. And so I go into FOIAbot and say, ‘I want to know how many accidents happened at First and Main.’ And FOIAbot identifies every single government agency that might have that information. With all of the formatting necessary for that particular government agency, it drafts the FOIA request, gives it to you, you sign it and it sends it. So we are able to send dozens of FOIAs a week.
And the reporters are the ones who are creating these. So Kristin, sitting at an executive committee meeting saying, ‘How should my team use it?’ That’s not the way to do it. It’s to say, ‘Here’s the tool. You tell me, help me help you. You tell me how you can use it.’ And that’s what’s working for us, democratizing access to the tool.
But with approved tools, where we have an enterprise agreement so that we are sure that the data is protected. You can’t go out freelancing with any tool you want. It’s got to be a tool that is given to us by a provider who has agreed to meet our very stringent standards around data protection. We have the highest standards for this. We’re a publicly traded company. And so any company that we do business with has to sign onto that, and we do not make exceptions when it comes to these AI. Then once they sign onto that, man, we push it to the people. We push it to the reporters and the editors and we say, let’s see what you can do with this.
Given the concerns about AI’s impact on journalism, how do you think about changing the narrative within the organization? How do you get reporters more interested in trying it?
Anton: We have town halls, where our own reporters, the ones that are the super users, share use cases. They talk about how they’re leveraging it. An example is our Taylor Swift reporter, Bryan West. You never know what Taylor’s going to do, so he may be out on the road covering something that’s not Taylor-related, and all of a sudden he’s going to go do the story, and he’s able to use AI. While he’s standing outside MSG, he’s using it to research stories that he’s already done to help him fact check something, or to create a listicle. So he loves it and he talks about it really enthusiastically. And that engenders trust amongst the rest of the team, to say, Bryan’s using it, then maybe I should try it for X, Y, Z. For [generating headlines], it took us a really long time to get everyone to say, This is actually great, that we have our own tool internally that gives you the key points at the top.
So I think it’s a matter of, internally, people need to start to just trust the process. And they’re hearing their colleagues use it. I think it has to be sort of organic. You can’t force it. We’re not perfect, but we’re certainly trying to bring everyone along on the journey, and create trainings so they can understand how to use it.
Reed: We’re also transparent about what we won’t use it for. Then you have to practice what you preach. If we did use AI for something we said we weren’t going to, we’d have no credibility. Over a longer period of time, us being transparent about what we’re not going to use it for and then operating the business that way we think will be helpful.
Are there things that you all like to use AI for, or that you find effective or that you’re hoping to use it for personally?
Reed: I use it personally just every day to help me. I use Claude and it knows me better and better all the time. I use it mostly in a business context, but whether it’s helping me create the right narrative for an email or helping me develop solutions for another business issue or something like that. It saves me a lot of time if I want to research something. How does this apply? Or what does this really mean? Or how are other companies using this? I use it for that. I think it helps me become smarter a lot quicker.
What sort of stance has the company taken on the use of its content for training and inference by the AI labs?
Anton: Our view is that high-quality journalism and content has real value and should be compensated when it is used in AI products. We have pursued commercial arrangements that recognize the value of our journalism, provide attribution, and create new revenue opportunities that support news gathering. The [now expired] Perplexity partnership was an example of that approach. It was built around making trusted journalism available within AI experiences while ensuring recognition for the original reporting. More broadly, we support responsible collaboration between publishers and AI companies rather than assuming those relationships need to be adversarial. We want to make deals happen. But they need to be fair.
Digital publishers once had lucrative deals with Facebook and YouTube that evaporated once the platforms had the content and audience.
Reed: Google does that all the time. They partner with you on business stuff and you build a business and as soon as they get it to where they want, they pull it up.
Anton: We built on our own platform with a partner, Taboola. We were the first to implement on USAToday.com, our own AI answer engine that only pulls answers—so if you go to a story, you’re reading about September 11th, in there you can ask it to go deeper—it’s called DeeperDive—on the story. So you know that within our wall of USA Today, those answers are coming from a trusted source. I think that’s a way of us saying, ‘Hey consumer, stay here longer, ask all the questions you want and know that you can rely on those answers.’ I think that matters. It’s sort of a learning. We’re adapting to understand, we need to keep you here. And you need to trust us, and you need to keep coming back to us, not to a random Google [search] that’s going to give you an answer. It’s not going to drive you back to us for more information.
When you look into the future—like five years—I wonder how you see newspapers, what their role is, how they fit into people’s daily lives. Does the newspaper as an idea change? Does the news business, does the media business have to change in some significant way or philosophically or structurally?
Reed: I think that what we’re changing is that news is one form of content and that we’re a content business. And so we create news and other things, whether it’s entertainment or sports or games or helping them.
But I think that in five years, I think the thing that’s continuing to evolve is we’re way more digital than we are currently. And in five years we’re going to be a lot more engaged with our consumers, because each one of them is going to feel like they’re getting what they need from us in a personal way. And so where we used to just push a set of stories into a market, we’re now going to be able to say, okay, we have this incredible pool of content. And for each user, pieces of that content are going to be relevant, and that’s what you’re going to be able to have. But we’re still going to be the news and media arm for consumers in local markets. But what’s evolving and changing is the personalized experience, and the fact that we’re a content business, and news is one form of content. There are many forms of content and we’re going to have to be good in all of this.
And ultimately people want, I think, to trust the news they’re getting. And there’s a huge dearth of trust right now.
Reed: I think that’s a huge asset that plays in our favor five years out. Whatever social platform you’re on, there are so many things that are not real. And it’s hard to know what to believe. So much stuff looks real. And so that’s the one asset we have that’s becoming a bigger and bigger asset, that trusted news and information and content, whatever we create. We create now beyond text. We have audio, we have video, but whatever we create, there’s a level of trust with it. And I think that’s a real differentiator.
And when I look out five years, I think one of the things that’s bringing consumers to us is not only that we’re unique from a local perspective, but that we’re a trusted source for all of the content you consume on our platform. I think that’s a differentiator five years from now. We’re going to become more valuable to the social media platforms too, because the owners of the social media platforms, they don’t want fake stuff. And so us creating content for those platforms I think is potentially not only a way for us to brand and bring consumers to us, but it’s a revenue stream from those platforms, because they’re going to want trusted content creators on their platform and we’re going to be high on that list.
I hope there’s still a print edition in five years too.
Reed: I don’t know if the frequency will be as much as it is today, but I think there’ll be print in five years. I really do. Our print revenue is 1.1 billion, and it’s not going to go away tomorrow. It’s declining at eight to 10%. But we still have an enormous amount of readers around the country who want print. What’s interesting is our print subscribers automatically get a digital code too, so they can be a digital subscriber. Less than a third of our print subscribers use digital. They only want print.
Correction: An earlier version of this story said USA Today Co. prohibits AI-generated stories. In fact, USA Today Co. policy permits AI-generated content with editorial oversight and disclosure.
