Previously in The Content Technologist:
Traffic for traffic's sake is a bad way to make money (2019). Minneapolis goes through it once (2020). Privacy is critical to remember when collecting audience data (2021). Minneapolis goes through it again (2026). We get angry. We cope. Our strong journalism and civically minded culture helps us orient and survive. At the Online News Association conference, news executives talk about AI (2026).
The frightening partnership of USA Today and Palantir
Journalists reporting on the massively successful analytics company Palantir tend to stick to its government contracts, which are frighteningly broad and opaque to the average American. Described this week in The Verge as "giant, data-sucking squid," Palantir's rapid expansion into American law enforcement and government operations has been cause for alarm.
What flies relatively under the radar is Palantir's growth into American private industry. While military contracts are the focus of most news investigations, Palantir's commercial contracts this year have nearly caught up with its government partnerships: per its quarterly earnings report, $764 million, or 48% of Q2 revenue, came from private-industry deals.
On August 6, USA Today Co. announced a partnership with Palantir in its quarterly earnings report, prompting a strong protest from the country's largest journalism union, the NewsGuild. Reporting from news trade publications has echoed the Guild's concerns about AI-generated text replacing workers, without specifically naming how Palantir's products would actually be used. On August 19, the Guild organized a minor walkout (a "coordinated lunch break," per the union's website) with the tagline "no slop in our shops."
Messaging from journalists and their union is relatively focused on the impact of genAI on the news product itself: the writing, the correctness and depth of its factual content, and its originality. There's an undertone of "the audience will lose trust in the journalism" because of LLM-generated text, and based on current popular opinion of AI from the people most likely to read news, that's a good prediction.
But the collaboration reflects a much bigger, more frightening spectre: Palantir, a private company willing to cooperate on building surveillance tools with the American government, likely has access to the personally identifiable information and news-reading behavior of—if USA Today Co.'s own reports are true—158 million average unique monthly visitors, or roughly 48% of the American people.*
What could Palantir do with that data? In their August 6 earnings call, USA Today Co. CEO Mike Reed said it will "strengthen how we collect, connect and activate audience data to drive more effective and faster monetization across our platform." I've sat through many ad tech, AI, and software demos over the past decade plus, and I've been an adviser on tactics like "more effective and faster monetization." As the call went on, Reed described the worst possible outcome for a news organization partnering with a surveillance-enablement company: "Converting the sheer scale of our audience into known, orchestrated, first party relationships."
To be clear: the CEO of the United States' largest news publisher is comfortable giving the known, orchestrated first-party relationships of its audience to Palantir, the company that built a tool used to target immigrant neighborhoods in Minnesota during Operation Metro Surge. That should raise red flags for media business wonks, journalism cheerleaders, and civil liberties warriors. We can't be so beaten down with doomscrolling or slop or whatever that we lose the freedom to access and understand news about our communities.
*USA Today's reported uniques likely aren't all from the US. In my experience media companies historically are hesitant or lacking the expertise to filter for duplicates or bots, but the numbers are from their official report, so who am I to question it? If they were to overstate their reach, I hear that would be securities fraud.
What does Palantir's Ontology product do?
In metaphysics, ontology is the study of being as it relates to the broader world. The term was introduced into the computer science community in the 80s and 90s to describe how AI systems could potentially relate concepts or entities. It's a central concept in Tim Berners-Lee et al's paper on the semantic web.
For the past few years, I've been steeped in semantic approaches to organizing data to be found by AI, and I've learned more about the construction of knowledge graphs, taxonomies, ontologies, and other semantic tools. These tools codify enterprise vocabularies and processes, applying language that connects business units and all manner of corporate operations. Many tech companies build their own taxonomies, ontologies, and knowledge graphs in-house, using semantic tools to audit and guide automated AI decision-making. Adding language and meaning guidelines to small language models can control hallucinations and reduce token expenses, making complex software more navigable through the use of an enterprise's its own preferred language.
Palantir's Foundry and Ontology products promise to help organize and analyze company data into a "dynamic foundation" to create "sovereign AI systems" per their website. They're sold to executives across the country who—rightfully—are frightened of providing their company's intellectual property and data wholesale to vendors like OpenAI and Anthropic. Executive buyers want shortcuts to efficiency and optimization without investing in building their own taxonomy and ontology software, a complex and time-consuming process that often involves months or years of deliberation from semantic geeks like me.
Palantir's off-the-shelf Ontology product is a language layer that enables access and organization of enterprise data without the arduous process of agreeing on a company vocabulary. Instead of owning their own enterprise vocabulary, companies can just use Palantir's language system instead. It would be a relatively innocuous product if it were created and sold by any other company.
For an executive frustrated that the promise of AI doesn't meet the reality, a product like Ontology is perceived as a turnkey solution to navigating the hidden secrets of their data. They're determined to find the magic solution without having to build it themselves. Deals with Palantir are happening now, across sectors, fueling the company's massive scale, and before long they will be irreversible. Corporate software procurement processes are long and arduous. Once software is integrated into an enterprise, it is extremely hard to extract from a company's operations, and that compounds when AI and semantic data are integrated. If Palantir data processing becomes as ubiquitous as Microsoft, Google, or Oracle in corporate software, it will take years to extract.
Pattern recognition: Media companies and privacy tech
Healthcare, finance, education, and credit bureaus have regulations on how users' private data can be processed, making their collaborations with Palantir somewhat less frightening. Palantir, at this point, can't violate HIPAA or FERPA without breaking the law. But there are no similar regulations controlling how media companies can process the immense amounts of personally identifying data. GDPR and CCPA cover limited regions and only regulate the specific case of selling user data to advertisers.
If this ill-considered collaboration were unprecedented, I would not be so concerned, but I've seen how vague language and publishers' misunderstanding of tracking software enabled the rise of Google's and Meta's ad networks. I advise clients toward software solutions and anonymized analytics approaches that respect individual privacy, and I've long observed that publishers take less care with audience and customer data than other industries. They litter their sites with the pixels of any upstart ad network that promises more revenue. They buy into tech like "identity resolution" with little trepidation or thought about how that software might compromise individual privacy, often while their reporters decry more consumer-facing surveillance tech like doorbell cameras and apps.
In the early 2010s, companies like USA Today, formerly Gannett, took Google and Facebook at their word and dropped third-party ad pixels on their websites, so they could siphon off a tiny portion of ad revenue while enabling Google and Meta to build audience behavioral interest graphs directly from those pixels. Those interest graphs are used to coordinate and sell advertising: "right message, right time, right audience" has long been a chorus in ad sales pitches for media and tech companies alike. But on the media side, collaboration with big tech only meant they lost their moat: their unique audience was no longer their audience. It morphed into ad tech's audience, where it remains.
USA Today Co. owns more than 100 U.S. news dailies, 1,000 weeklies, and is actively buying up struggling local news operations. A USA Today-owned paper is the only local news source in many U.S. communities. It runs the paper in my hometown, and there's a good chance it runs yours. It operates papers in cities like El Paso, Texas; Alexandria and Shreveport, Louisiana; and Fort Collins, Colorado; all of which happen to be in close proximity to ICE detention centers.
Trusted brands give the news giant access to local audiences, who don't pay much attention to ownership structures or the news business. They're there to read about what's going on in their communities. To access and pay for their news, they have to submit their email address and zip code information. They will never know they're submitting that personally identifiable data to be processed by Palantir, just like they didn't know that by reading the news they were giving their anonymized behavioral patterns to Google and Facebook ten years ago. They are just reading the news.
Enterprise legacy news companies are years behind on digital transformation compared with other Fortune 500s. USA Today Co. is just now realizing that quality-over-quantity might be the better move, after years of belatedly chasing lowest common denominator SEO terms. Their ham-fisted partnerships with ad tech companies have brought us innovations like chum and the "news" website whose per-page content/ad balance is somewhere around 15/85. They don't have a good history of protecting, or even acknowledging, their audience's privacy. Instead of building decent, uncluttered products, they will sell their audience's behavioral data to whichever third-party solution makes the best pitch.
Do you trust Alex Karp and Peter Thiel?
When working in corporate content marketing, clients generally want their websites and communications to appear benign and safe. They don't wish to attract controversy, so their product copy is banal and innocuous. Not so with Palantir, whose website contains some of the most wackadoo** phrases I've ever seen on a corporate site, like "Govern the human-agent labor force" and "AI sovereignty is your alpha." Its marketing promises that customers will be free from the whims of big tech pricing and technical changes, while selling clients their proprietary systems.
Their target audience is, weirdly, "commercial enterprises in The West," a segment that I've never seen named before in all my years of working with software and marketing.† Co-founder Peter Thiel is, historically, no friend of journalism and in years past, publicly made anti-democratic statements. CEO Alex Karp has a habit of dropping megalomaniacal political statements into Palantir corporate social media and shareholder letters. Their company is named after a magical evil surveillance tool from Lord of the Rings. In previous years, I would find the confluence of villain vibes laughable. In 2026, the potential for harm keeps me up at night.
"We stand on the walls, sentinels of the inner sanctum, against the assault of AI slop. The Ontology is based firmly in reality—there is, here, a dialectic between ground truth, tribal knowledge, and enhancements."– Alex Karp, Palantir Q1 2026 shareholder letter.
[I went to a conference for ontologists this year, and I assure you: most semantic professionals do not sound anything like this guy.]
Unlike most MBA-trained CEOs, Karp is extremely aware of the power of language to communicate ideology and start political movements. Recently voted the least trustworthy AI leader in a CNBC survey of Gen Z respondents with a whopping 19% trustworthiness approval rating, Karp positions the Ontology product as an answer to "AI Slop," a narrative conveniently aligned with the popular backlash against language generators. That his company's biggest U.S. commercial product is a language-control tool should raise red flags—but tech and culture reporters seem to be more interested in Palantir's branded merchandise than the potential of its software to define and control the vocabulary of American business and government.
Linguistic and technology behavioral shifts happen gradually, incrementally, over five or ten years. If a company were to optimize toward readability from Palantir's Ontology product, it might change its language to reflect the words Palantir is already using. For example, in ten years, a consumer may be buying from an Ontology-powered home insurer. After requesting a quote, the customers who self-identify as being from The West are treated to lower rates than those not from The West. Hypothetically. With precedent.
American media companies are increasingly eyeing revenue-generation software that relies on surveillance pricing, adjusting subscription offers based on a referral link's or customer id's predicted income. They often don't understand how the software works before signing the contract, then panic when the algorithms change. Publishers panic loudly at the loss of traffic-based ad and affiliate revenue from Google, drowning out conversation about building better products or exploring meaningful alternatives. They're happy to align with another company that ensures that their intellectual property is "sovereign," audience privacy be damned.
**"Wackadoo" is a work-friendly Midwestern stand-in for a string of curses.
†The early aughts freshman curriculum at NYU taught that "The West" or "western civ" was a meaningless term in a global society. Perhaps I was subject to liberal bias or indoctrination from wildly Marxist professors, but I am now a fully grown adult, and I can decide for myself that The West is a silly label. That descriptions of "Westerners" often seem to exclude Africa and the Americas south of Texas, or anyone who is not Judeo-Christian, renders it a meaningless category.
"Nobody talks about Palantir."
Late last month I had lunch with another consultant in New York, who lists many major media companies as clients. We talked about how AI-first products might work with news consumers, and I gave them a rundown of what I'd heard at ONA. Toward the end of the conversation, I mentioned knowledge graphs and ontologies as a means to build better information products. I brought up the Palantir Ontology project, completely clueless about their previous deals with Fox News and Axel Springer. Was Palantir an emerging character in the news revenue scramble? I wondered. They shook their head: "Nobody talks about Palantir."‡
I didn't perceive their line delivery as sinister in the moment, but after I read the USA Today announcement, I heard the (probably unintentional) subtext. The business and legal side of the news industry would prefer not to tangle with the Big Bad that killed Gawker ten years ago.
This essay has been sitting in my gut for two weeks, and I haven't seen any reporters write about the deal from the audience privacy perspective. I don't want to be the Cassandra in the content mines. As a consultant, I'd rather build good systems than call out the bad ones, and I feel like a conspiracy theorist writing it out. But as a citizen whose perception of country and freedom keeps getting ripped into tinier and tinier pieces, and as an experienced professional who knows how "orchestrating personal data" works, I'm scared.
I am not a lawyer, but I advise companies to talk to their lawyers when they procure audience data-processing software or make changes to their privacy policy. From where I'm sitting, the best case scenario in the partnership is that Palantir helps USA Today develop a new iteration of the chum bucket.
In its current unregulated state, the risks of a massive news organization working with a politically motivated organization like Palantir are extremely high: Palantir could build its own interest graph from USA Today's data, which could be used to provide information to law enforcement on zip codes that index high for interest in stories about protest actions. Stories about immigration regulation and the abuses of ICE could be removed or deprioritized from news websites if they are not generating enough revenue. Individuals who seek out those news stories could be identified. Palantir may develop political advertising products similar to Meta's, like those described in Sarah Wynn-Williams' Careless People. Instead of laws, all that protects the audience are Palantir's opaque corporate policies and the word of its co-founders that those protections won't suddenly change on a whim.
Media businesses are struggling, but that doesn't mean they have the right to compromise their audience's privacy and, potentially, their safety and autonomy. The capacity for more than 100 million people to access the news privately and confidentially is now at the whim of a proudly military-aligned data processor that cares more about protecting corporate sovereignty than human freedom and individual autonomy.
‡Normally I keep business lunches completely private, but this particular consultant converted our conversation into LinkedIn posts less than an hour later, so I feel no guilt in sharing my direct experience.
Want to do something about it?
- If you're a journalist, please cover this story in more depth. The above is all desk research combined with some professional knowledge.
- If you buy advertising, ask your vendors whether they work with partners who collaborate with Palantir. Simply asking is often enough to raise alarm.
- If you live in a community with a USA Today Co. paper, call them (subscription lines, not editorial) and ask them what's up with your privacy as an audience member.
- Support the NewsGuild, who raised the alarm. (I am not a NewsGuild member, just a fan.)
- Spread the word. I don't care whether you link me or not. Just stop doomscrolling and talk about it with your friends and colleagues. I fully believe in the power of speech to make real change.
- Understand what AI-integrated software can do now, and focus on what you can solve in the present tense. Look at quarterly earnings and ongoing deals in addition to data center plans. The most pressing concerns have nothing to do with the consumer layer of the internet: slop, branded merchandise, em dashes, fiction. There's no money in that. There's money in software. Businesses make bad tech decisions when they fear they'll be left behind, and they're making those deals now.
The Content Technologist is a newsletter and consultancy based in Minneapolis, working with clients and collaborators around the world. The entire newsletter is written and edited by Deborah Carver, an independent content strategy consultant.
Affiliate referrals: Ghost publishing system
Did you read? is the assorted content at the very bottom of the email. Cultural recommendations, off-kilter thoughts, and quotes from foundational works of media theory we first read in college—all fair game for this section.