Axios sells three years of training rights to test how little newsroom it needs

OpenAI is covering the startup costs, staff and technology included, of 13 new Axios Local newsletters, and in exchange Axios is letting OpenAI train its models on its free published coverage for three years ^[1][3]. Axios employees also get free credits for OpenAI's enterprise tools ^[2]. The arrangement was set out by Issie Lapowsky for the Columbia Journalism Review, and it is the fullest account of the deal so far ^[4]. What is being tested is narrower and more consequential than a content licence. Jim VandeHei, the chief executive, framed it to CJR as a question: "Could we use AI to basically automate everything you need in a local market other than the journalist and the journalism itself?" ^[5] That is a staffing hypothesis dressed as a technology one. If it holds, the reference cost of a local newsroom drops to reporters plus a model subscription, and every other content business gets offered the same trade. The timing sits oddly with its own author. On 9 August, VandeHei published a column arguing that change now outruns anyone's ability to absorb it, and wrote that "for the first time ever, my confident line of sight is six weeks tops" ^[6]. The deal he signed runs three years, roughly 26 times that horizon, and asks reporters in 13 markets to build careers inside it [c7a]^[3]. The undisclosed parts are the load-bearing ones. Allison Murphy, Axios's chief operating officer, told CJR that OpenAI invested enough to cover the startup of the new markets, and neither company would name the figure ^[7]. Neither would say whether the deal bars Axios from taking legal action against OpenAI; an OpenAI spokesperson said the terms are confidential ^[8]. Local newspapers are currently suing OpenAI over training on their work ^[9]. Without that answer, an operator cannot tell whether this is a licensing deal or a pre-emptive settlement, and those two things have very different price tags. The distribution of the money is also worth reading closely. OpenAI funded four starter cities first, all chosen by Axios: Pittsburgh, Kansas City, Boulder and Huntsville, of which only Huntsville lacks a daily newspaper and none of the others would be classified as a news desert ^[11]. VandeHei called covering communities with no news at all "the Holy Grail" but said Axios has to prove the economics first ^[12]. Axios Local has more than two million subscribers and does not make money; a spokesperson said that was expected, and VandeHei said each city turning a profit within five years would count as a great success ^[13]. Matt Pearce, policy director at Rebuild Local News, a coalition representing more than 3,000 newsrooms, told CJR that "if you strike a direct partnership with an AI company, that benefits you but no one else" ^[14]. Internally, the leverage is one-sided. Axios cut 10% of its staff months before the deal and another 11 jobs this year, and is not unionised ^[15]. At Politico, which VandeHei co-founded, the union won an arbitration that shut down two AI tools ^[16]. On an OpenAI panel in March, VandeHei said Axios was "early to basically telling our staff that you don't have a choice if you're scared, if you don't like it, we don't care" ^[17]. Holly Moore, executive editor of Axios Local, told CJR that worries eased once it was clear nobody had to use AI in reporting ^[18]. The Boston Globe's Shira Center, on the same panel, said her paper has no intention of replacing reporters with bots ^[19]. The counter-case is a person. Robert Sanchez writes two Axios newsletters covering Douglas and Arapahoe counties in Colorado, an area larger than Rhode Island, and every local byline is his ^[20]. Whether that is coverage or a coverage-shaped hole depends on what the automation actually absorbs.