Karen Hao had already booked her tickets to San Francisco. OpenAI had told her she could come back to its office and sit down for interviews for her book. Then, she said, the visit kept getting pushed back. Finally an email arrived: "we are not going to participate at all."
Hao is a journalist who covers AI's effect on society. She wrote Empire of AI, an investigation of OpenAI, the company behind ChatGPT. On The Diary of a CEO, she told host Steven Bartlett how her relationship with OpenAI went over several years, and how she came to see being shut out as an advantage. The episode, published October 23, 2026, replays a moment from an earlier full conversation with her.
At issue is access. Reporters who cover tech companies depend on them for interviews, office visits and responses to a "request for comment," the standard step where a journalist sends a company the claims they plan to publish and asks for its side. A company can grant or withhold all of those.
A profile leadership disliked
Hao said the trouble began with a profile of OpenAI for MIT Technology Review. She spent three days embedded in the company's office in 2019, and the piece came out in 2020. "The leadership are very unhappy," she said. According to Hao, her book quotes an email that OpenAI chief executive Sam Altman sent to the company about the profile, which she summed up as saying, "this is not great."
The friction started during the reporting. The Italian edition of the 2020 profile describes limits on her access to the office and a rule that her interview requests needed the communications team's approval. Hao objected to that rule, arguing that approval could make sources less candid.
After the profile came out, Hao said, OpenAI's position was that it would not take part in anything she did or answer any of her questions. She stressed that the company said this outright: "It wasn't like me inferring." Her example was a colleague at MIT Technology Review who also covered AI. OpenAI sent him a press release, and when he said he was busy and asked, "Will you send it to Karen?", the reply was, "oh no, we have a history you understand."
Three years of silence, then a thaw
Hao said the company refused to talk to her for about three years. That changed when she moved to the Wall Street Journal, where she believes OpenAI "felt a bit compelled because it was the journal" to reopen contact. She sent requests for comment on every piece and asked for interviews, and she said the two sides reached "a more productive relationship."
When she left the Journal to write the book full time, she told OpenAI right away and said she wanted to make sure the book reflected the company's perspective. In her telling, OpenAI agreed: "we can arrange interviews for you. You can come back to the office."
The firing and 40 pages of questions
While they were still arranging the visit, OpenAI's board fired Altman, the 2023 episode that ended with his dismissal and return. Hao said that was when "things started going kind of south," because the company became "very sensitive to scrutiny." The scheduling kept slipping, which she described as "kicking the can down the road," until the email arrived calling everything off after she had already booked her flights.
She told OpenAI she would still give it a chance to respond. She kept the company updated on what her reporting found, sent 40 pages of requests for comment and gave it more than a month to reply. OpenAI, she said, never responded to a single one of the 40 pages.
A tweet about two books
That back-and-forth was still going on, Hao said, when Altman posted about the books coming out on him and OpenAI. As Bartlett read it on the show, Altman wrote that the company had taken part in only two: one by Keach Hagey focused on him, and one by Ashlee Vance on OpenAI. He added that "no book will get everything right, especially when some people are so intent on twisting things." Hao quote-tweeted it to say that the unnamed book was hers.
Asked whether Altman meant her book, Hao answered: "100%. Because there's only three books coming out about him." Hagey's biography, The Optimist, came out on May 20, 2025, the same day as Empire of AI. Vance's company, Core Memory, lists an AI and OpenAI book among its development projects.
Access as a carrot
Bartlett then turned to his own show. Altman does many interviews, he said, from Tucker Carlson (Altman's episode on that show was published September 10, 2025) to Theo Von and Joe Rogan. "I wonder why he won't do mine," Bartlett said. He described himself as fair with guests and said he had "heard through the grapevine" that Altman doesn't want to come on.
Hao said companies treat anyone with a large platform the way they treat journalists. "It's not just about the conversation that you're going to have with them," she said. "It's about who you also choose to platform." In her view, companies know that "a really big carrot that they can give to technology journalists is access," and they will withhold it "at the drop of a hat" if they learn a reporter is talking to someone they didn't want them to talk to.
Bartlett said this matched his experience. Without naming anyone, he said the team of "a particular person in AI" had "dangled the carrot" of an appearance on his show for about 18 months, with messages along the lines of: he's thinking about it, let's think about a date. Bartlett said he reads it as a strategy: if the invitation hangs there long enough, the host stays pleasant, avoids criticism and keeps the company's critics off the show. "I think a lot of their game is just dangle the carrot forever," he said.
He said the tactic won't work on him. He will talk to whoever he wants, and if the person does come, he said he will "give them a fair shot" and ask genuinely curious questions about their work and incentives: "I won't gotcha them." He argued that public figures now have to show up, give their opinion and allow "the clash of ideas in the public forum," so viewers can decide for themselves.
Why she now calls it lucky
Hao called these tactics "such a huge part of their machinery" for managing the companies' public image and controlling which information and opinions get out. Looking back, she said, "I feel very lucky now" that OpenAI shut her out early.
She didn't feel that way when it happened. She was "relatively junior" in her career, she said, and "felt like I had screwed myself over." She wondered whether she should have been nicer in the profile to keep her access, which she called "a horrible question to ask as a journalist," because the job is to report the truth in the public's interest. She also asked herself whether she should have been "playing the access game." But by then the door had already closed, and she built her career knowing "the front door was never going to be open."
Hao said that knowledge "really strengthened my own ability to just tell it like it is": to report the facts as she found them, whether the company liked it or not. Most often, she said, it does not. "They don't need to open the front door for me," Hao said. "I was still able to do more than 300 interviews." Earlier in the conversation, Bartlett had put the total at about 300, 80 or 90 of them with people from OpenAI.