Politico became one of the first newsrooms last year to win a union contract that included rules on how the media outlet can deploy artificial intelligence. The PEN Guild, which represents Politico and its sister publication, environment and energy site E&E News, is now gearing up for another first. The union’s members allege that the AI provisions in their contract have been violated, and they’re preparing for a groundbreaking legal dispute with management. The outcome could set a precedent for how much input journalists ultimately have over how AI is used in their newsrooms.
Last year, Politico began publishing AI-generated live news summaries during big political events like the Democratic National Convention and the US vice presidential debates. This March, it debuted a suite of AI tools called Policy Intelligence Assistance for paying subscribers, which were built in partnership with the Y Combinator-backed startup Capitol AI. Politico executive Rachel Loeffler described the initiative at the time as “seamlessly integrating generative AI with our unmatched policy expertise.”
Politico union members, however, allege these tools violated their contract in several ways, and are taking the dispute to arbitration this July. “The company is required to give us 60 days notice of any use of new technology that will materially and substantively impact bargaining unit job duties,” says PEN union chair and E&E public health reporter Ariel Wittenberg. The union claims that it was given neither notice nor an opportunity to bargain in good faith over Politico’s AI rollout, and that the tools do work that would ordinarily be done by human staff.
“This isn’t just a contract dispute, it’s a test of whether journalists have a say in how AI is used in our work. With no federal rules in place, union contracts remain one of the only enforceable frameworks for AI accountability on a national scale,” says Newsguild president Jon Schleuss. (PEN Guild is a unit within Newsguild, one of the most prominent unions for journalists.)
Politico says the publication “takes the obligations under its collective bargaining agreement seriously,” and “will continue to honor those obligations while also rapidly embracing transformative technologies such as AI that will revolutionize how our audience consumes news and information,” according to spokesperson Heather Riley.
Politico’s contract stipulates that the publication needs to use AI in a manner that follows the company’s standards of journalistic ethics. “We’re not against AI, but it should be held to the same ethical and style standards as our political journalists,” says Arianna Skibell, the union’s vice chair for contract enforcement, who writes Politico’s energy industry newsletter. Some union members question whether there’s always appropriate human oversight over the AI content Politico publishes.
In one case, an AI-generated live summary used language around immigration that human writers are not permitted to use, publishing phrases like “criminal migrants” as it covered the vice presidential debates.
“There were also factual errors that the AI inserted that night,” alleges Skibell. For example, she says, the AI credited actions taken by the Biden Administration as things Kamala Harris did. That post was later swapped for replacements without the errors, according to screenshots reviewed by WIRED. “At Politico, you can’t just wholly take down articles written by human reporters without going through a series of approvals, all the way up to newsroom leadership. That did not happen for the AI live summaries,” Wittenberg claims. (Politico did not comment on the specifics of the union’s allegations.)
Union members say they believe the AI-generated posts were handled in a way that violates Politico’s correction and takedown policies. They also allege that one of Politico’s paid premium AI tools for generating policy reports has spewed out incorrect information in the past. Politico’s human reporters broke the news in 2022 that the Supreme Court had voted to overturn Roe v. Wade, but a report on abortion rights generated by the Policy Intelligence Assistance tool in March 2025 was written as though the constitutional right was still in effect.
Politico’s experiments with AI are not unique in the journalism industry. The publication’s parent company, German media conglomerate Axel Springer, is one of a number of news giants that have entered into multiyear licensing deals with OpenAI, including Vox Media, The Atlantic, and WIRED’s parent company Condé Nast.
Artificial intelligence tools have been leveraged by journalists to produce award-winning work. Two recent Pulitzer prize winners, for example, disclosed using machine learning and AI in their reporting process. In other cases, though, AI features have gone awry, like when tech news site CNET published error-addled AI-generated articles that were deemed a “journalistic disaster” by The Washington Post.
As media companies race to integrate even more powerful and pervasive AI tools into newsrooms, sometimes in direct collaboration with tech companies, the writers and reporters on staff frequently have little say over how the technology is adopted. Politico’s fight could help change that by setting a precedent across the industry. “AI decisions have been few and far between,” says labor lawyer Alek Felstiner. “That would magnify the importance of an arbitrator’s decision beyond what it would normally carry, because it’s going to be one of the first observed on a critical issue.”
Although this is the first clash of its kind in digital media, the entertainment industry has fought similar battles in recent years as creatives chafed against certain types of AI usage. In 2023, two of Hollywood’s biggest unions—the Screen Actor’s Guild and the Writer’s Guild of America West—won hotly contested contracts that included AI protections. This week, SAG-AFTRA filed an unfair labor practice charge against a production company associated with the video game Fortnite for using an AI-generated replica of late actor James Earl Jones’ Darth Vader voice.
“We do remain hopeful that we can come to some kind of agreement,” says Skibell. “But we’re also ready for a fight.”
