Sign up for the daily CJR newsletter.

Robert Sanchez, a forty-nine-year-old journalist in Colorado, started working in local news before graduating high school—inspired from a young age by Beckett Baseball Card Monthly. At his local weekly paper in Parker, Colorado, he began a type of career that’s rare these days: he covered sports and town hall, then moved to bigger publications. It’s been a gratifying experience. “I love my wife and my kids and journalism,” he told me. And since he and his wife worked on the high school newspaper together, in a way, he owes those first two loves to the third. 

But he has also felt the industry sink, dodging layoffs and weathering buyouts. The arrival of AI has brought new worries. Then, early this year, he spotted a promising job listing.

Axios was hiring a writer for two newsletters—one in Sanchez’s own Douglas County, Colorado, and another in neighboring Arapahoe County. The gig was funded by a new partnership between Axios and OpenAI. Under a three-year deal, OpenAI would bankroll the startup of thirteen Axios Local newsletters and give Axios staff free credits to access its enterprise-grade tools, while Axios, in turn, would allow OpenAI to train its models on Axios’s free, published coverage. 

Axios viewed the deal as a chance to test whether leaner local newsrooms, armed with all of AI’s efficiencies, could be economically viable, or even a growing business. As Jim VandeHei, the cofounder and CEO of Axios, put it to me: “Could we use AI to basically automate everything you need in a local market other than the journalist and the journalism itself?” Critics saw it as a devil’s bargain, enabling an 852-billion-dollar Silicon Valley giant to choose local journalism’s winners and losers while burnishing its own ailing reputation. To Sanchez, it looked like something else: a life raft in an industry being rapidly pulled under.

Every media company faces existential questions about the rise of AI, especially this one: How should newsrooms get paid for the information AI spits out and the models that were trained on journalists’ work? The New York Times has sought to answer this question in court. Some (including the Times) have struck licensing deals. But as most newsrooms scramble to limit AI’s damage, VandeHei has been among the loudest voices arguing it may well do some good. “AI will save local news,” he declared during a panel hosted by OpenAI in March. 

VandeHei, who has described himself as an “AI lab rat,” uses the tools obsessively, talking to Claude on long hikes (as he trains for climbing Mount Kilimanjaro) and jumping out of bed at 3am “jazzed to test or explore a new idea.” He has argued that anyone who sees AI as overhyped is “on drugs or just not using it correctly” and recently declared that AI is “a better editor and a better writer than anyone at the company.”

VandeHei’s boosterism has invited no shortage of skepticism. “Is anyone else starting to worry about @JimVandeHei?” a Variety media reporter wrote on X in June.

But on a recent morning in Axios’s Arlington, Virginia, offices, VandeHei appeared unapologetic about his embrace of AI. “We try not to get distracted by what other people in the quote-unquote media think,” he told me over Zoom, while his cofounder Mike Allen sat just out of frame. “I’m not delusional,” he insisted. He worries about media companies refusing to reckon with historic change. “I just don’t have a whole lot of time for people who are AI deniers,” he said. “Like, good luck.”

At fifty-five, VandeHei presents with the swagger of a person who has been right before. In 2006, when he and John Harris left their jobs at the Washington Post to cofound Politico with Allen and Robert Allbritton, the media mogul, it was based on an (accurate) assumption that the internet would reward a more propulsive, insidery brand of political journalism. In 2017, VandeHei and his cofounders did it again with Axios, betting that, amid dwindling attention spans and digital advertising arbitrage, bite-size newsletters could overcome what he called the “crap trap” of the open Web. By 2020, local news seemed like the next opportunity. VandeHei’s thinking went like this: Local newsrooms of old had high fixed costs. Buildings. Paper. All of it became dead weight as revenue plummeted—and none of those costs had much to do with reporting. “If your cost was essentially the journalism, that’s relatively cheap,” he said.

In December of 2020, Axios acquired a North Carolina newsletter called the Charlotte Agenda. Within a few months, Axios Local had hired journalists in five additional markets, with another eight newsletters launching by the end of 2021. These newsletters covered a range of beats, including election coverage, city council votes, and restaurant trends. Each Axios Local newsroom consists of one to four journalists working with an Axios Local editor who covers multiple areas. Every reporter works remotely within their market.

Even in the early days, VandeHei was already dreaming up what he called a “local supersystem” that could automate some rote tasks. Then, in August of 2022, Axios was sold to Cox Enterprises in a deal that reportedly valued the company at 525 million dollars, and came with a twenty-five-million-dollar investment to expand both local and national coverage. Three months after the sale, ChatGPT made its public debut.

If the supersystem were powered by something like ChatGPT, VandeHei told me, he believed that could give reporters more time to “do what only humans can do, which is really awesome reporting.” He imagined partnering with an AI company. In late 2023, he began having conversations with several of the most prominent AI companies and their leaders, including Sam Altman, the CEO of OpenAI, about the potential impact AI would have on the media. From the start, Allison Murphy, the chief operating officer of Axios, said, the company wanted more out of OpenAI than to simply get paid for training data. “It was about local,” she told me. “It was built around our shared belief that a combination of expert human reporters and technology could be a solution—not the only—but a solution to some of the needs for local news.”

OpenAI jumped at the idea. In particular, VandeHei said, the proposal won over Chris Lehane, OpenAI’s chief of global affairs and a longtime Democratic operative dating back to the Clinton White House, who has lately sought to sway popular opinion about AI through a number of civic-minded initiatives. OpenAI had already offered grants, content-licensing agreements, and technical capabilities to a number of media organizations, including the American Journalism Project and the World Association of News Publishers. But the Axios proposal was the first partnership that attempted to actually fund new newsrooms. “We thought that this partnership was a really strong example of AI being able to create new opportunities where they hadn’t existed before,” Varun Shetty, the vice president of media at OpenAI, told me.

VandeHei put it more bluntly: “They need some wins.”

OpenAI also needed the information local news produces. As of February, ChatGPT users in the United States were writing a million prompts per week seeking out local news, according to the company. But dwindling local news risks creating a dearth of information for AI tools to pull from. “ChatGPT can help people find local news,” OpenAI recently wrote, “provided local journalists are still out there covering it.” 

VandeHei debuted the OpenAI deal in January of 2025 with a characteristically succinct post announcing that, for three years, OpenAI would fund four starter cities—Pittsburgh; Kansas City, Missouri; Boulder, Colorado; and Huntsville, Alabama. All of them were Axios’s picks. This year, the two companies added nine more newsletters, adding to Axios’s expanding roster. By the end of 2026, Axios is slated to be in forty-three cities, and VandeHei told me he would like to someday be in a thousand. “We want to be in every community, big and small,” he said.

Murphy, Axios’s COO, said that OpenAI invested enough to cover the startup of the new markets, including personnel and technology. But neither Axios nor OpenAI would disclose additional details of the deal, including the exact amount OpenAI is paying or whether the deal precludes Axios from taking legal action against OpenAI. Kayla Wood, an OpenAI spokesperson, said these terms are confidential under the agreement.

As of early August, Axios’s open local job listings advertised salaries from sixty-five thousand to a hundred and twenty-five thousand dollars. Axios doesn’t disclose its OpenAI deal in those job listings or in most of its coverage about OpenAI, unless the coverage explicitly relates to media deals. 

An Axios staffer who spoke on the condition of anonymity said the lack of transparency has contributed to some unease among rank-and-file reporters. Holly Moore, the executive editor of Axios Local, acknowledges there were questions early on about whether it was wise to let OpenAI train on stories. “Once we made it super clear that there was no requirement that we were using AI as part of the reporting process,” she said, “then people started being like, ‘Oh, okay.’ It’s very much a business partnership.”

Then again, what other option did they have? Journalism’s deep cuts haven’t spared Axios, which laid off 10 percent of its staff months before the OpenAI deal and this year cut another eleven jobs across the news desk and visual and social media departments. Unlike other newsrooms, such as Politico, that have successfully fought back against controversial uses of AI, Axios is not unionized. And VandeHei has hardly been shy about the need for employees to get on board. “We were early to basically telling our staff that you don’t have a choice,” he said during the panel hosted by OpenAI. “If you’re scared, if you don’t like it, we don’t care—you have to embrace it.”

When I asked VandeHei about the potential for conflicts of interest—or even the perception of conflicts, since Axios regularly covers OpenAI—he appeared unfazed. Axios’s deal with OpenAI is “silent” on the subject of the editorial firewall, he said, not because it’s unimportant, but because both sides took Axios’s editorial independence as a given. “There’s an assumption that we’re gonna write what we write,” he said. Axios has continually covered damning allegations against OpenAI over ChatGPT’s alleged role in teen suicides and each beat of OpenAI’s ugly court battle with Elon Musk. While much of this coverage is handled by the national tech desk, local reporters in the cities OpenAI helped fund have covered data center fights in their communities.

VandeHei argued that it’s up to readers to determine if Axios’s coverage is biased in favor of OpenAI or any other sponsor. “The market will set the relationship,” he said.

But some have raised concerns about the Axios Local model’s true impact on the industry. Matt Pearce, the director of policy for Rebuild Local News, a coalition of over fifty-five organizations that represent more than three thousand newsrooms, pointed out the “collective action” problem for newsrooms, where it is often more advantageous to strike deals with partners like OpenAI than to take them to court. And yet, he said, “If you strike a direct partnership with an AI company, that benefits you but no one else. There’s a systemic problem, and you’re not really doing anything to address the systemic problem. If anything, you’re kind of benefiting from it.” 

With the possible exception of Huntsville, which has no daily newspaper, none of the starter cities Axios chose are what researchers tend to classify as news deserts. VandeHei told me that serving communities with no news coverage would be “the Holy Grail,” but Axios still needs to prove the economics can work in established markets. Though Axios Local now has more than two million subscribers, it is not profitable. (An Axios spokesperson said this was expected. VandeHei told me that if each city turned a profit within five years, it would be “a great success.”)

VandeHei agrees that both local and national newsrooms need more than one-off investments from AI companies. “I sure as hell hope the answer is that there’s a compensation mechanism,” he said. Maybe that gets decided by Congress. Maybe it comes from the courts. But as the head of a comparatively young news organization that already gives most of its stories away for free, VandeHei told me, he didn’t believe Axios was well-positioned to fight that particular battle.

The OpenAI deal was, he said, a financial transaction. “At the end of the day, if someone’s going to pay you for content, you say: Well, how much are you going to pay me?” he said. “If that payment feels like it’s a fair transaction price, you say yes.”

Back in Colorado, Sanchez said he was drawn to the way Axios was thinking about AI. “I want to be one of the people who’s not sitting on my hands like I saw everybody doing when I first started in this business,” he said, recalling the early days of the internet. 

Sanchez is now working from home, publishing the Arapahoe County newsletter on Mondays and Wednesdays and the Douglas County newsletter on Tuesdays and Thursdays. All of the local bylines are his. This has been a tremendous task: together, the two regions constitute a geographic area larger than Rhode Island. Competing outlets already have multiple reporters assigned to cities on his beat. “So there are all these built-in challenges,” he said.

But he’s also begun to see how leaning on AI can help meet those challenges. Recently, while researching a story on AI in policing, he used ChatGPT to help draft a document request under Colorado’s Open Records Act and find the right contact for it, and then he ran it through an automated legal filter Axios built to minimize the odds it would be rejected. 

Once he got the trove of documents back, he ran it through ChatGPT again to see whether the agency had produced all of the records he asked for. It hadn’t. He had ChatGPT draft another letter saying as much. By the end of the day, he had the three missing files in hand.

Editor’s Note: This article has been updated to clarify a quote by Matt Pearce. 

This piece was produced with support from the Craig Newmark Center for Journalism Ethics and Security.

Has America ever needed a media defender more than now? Help us by joining CJR today.