The Viral AI Warning Wasn’t as Spontaneous as It Looked
Jacob Coxon told Fox nobody helped his rollout. He’d already told the Wall Street Journal that a friend did, and a small circle moved within minutes of his post.
By Lynn Matthews - September 25, 2026
Jacob Coxon quit Anthropic on the evening of Tuesday, Sept. 8, and posted an X thread warning that artificial intelligence could kill humanity by the end of the decade. He had fewer than 100 followers at the time. The thread went on to draw nearly 174 million views, and within days he was on every major network. The next day, on Fox News, he was asked whether anyone helped him plan the rollout and said no. He had already told a different story to the Wall Street Journal.
What Coxon Told the Wall Street Journal
According to the Journal’s Sept. 18 profile of Coxon, he told his Anthropic colleagues he was quitting in a Slack message that Tuesday afternoon. About 30 minutes later, the Journal reported the news. Coxon then posted his resignation thread from a park bench in San Francisco.
Before any of that, Coxon said, he “turned to his friend Nathan Calvin for help” announcing his exit. Calvin is general counsel of Encode AI, an AI safety advocacy nonprofit. Within minutes, Calvin, former OpenAI safety researcher Daniel Kokotajlo, and other AI safety advocates, in the Journal’s words “including many of his former colleagues,” shared it. Coxon has also described setting up a small group chat, about 10 people, to help push the thread out early. That night, Anthropic’s own alignment lead, Evan Hubinger, publicly backed Coxon, writing that he believed the odds of AI killing everyone within a decade topped 10 percent.
The Journal’s own reporting names the dispute this created. Its piece notes that AI boosters view Coxon’s exit as part of what they call a coordinated plot to push through stronger global regulation, one they say has been promoted by figures in the effective-altruism movement.
What He Told Fox News
The next day, Wednesday, Sept. 9, Fox News’ Bret Baier asked Coxon on air: “Do you work with any third parties as far as this whistleblowing and coming out?” Coxon answered, “Not at all. This is my own personal concerns about safety and conversations I have had.” The exchange is on the record on video. Pirate Wires flagged it as part of a wider story on his media rollout. Pirate Wires is an outlet that generally favors faster AI development and refers to Coxon and his allies as “doomers,” worth knowing as the frame the story arrives in.
Baier introduced the segment by thanking Coxon for “telling it here first on ‘Special Report,’” framing it as an exclusive first broadcast appearance. That sits oddly next to the fact that the Wall Street Journal had already run its own exclusive interview with Coxon roughly 30 minutes before his thread even posted the day before.
Coxon’s answer to Baier left out Calvin, whose help with the announcement Coxon had already put on the record with the Journal.
A Second Push, After the Fact
Pirate Wires, citing two sources and an internal email, reported that a communications firm, DEY. Ideas + Influence, was separately working to book Coxon media interviews around the same period. The outlet also reported that DEY. had arranged interviews for Nate Soares, president of the Machine Intelligence Research Institute, timed to overlap with Coxon’s public debut.
Soares confirmed the outreach himself, in a post on X quoted by another user after the Pirate Wires story ran. He wrote that once Coxon’s thread passed roughly 100,000 likes, people in the AI safety community started pressing him to get Coxon more airtime, so he asked DEY. to help. By the time the post passed 500,000 likes, Soares wrote, DEY. had landed exactly zero hits.
Neither Coxon nor DEY. responded to Pirate Wires’ requests for comment.
Who DEY. Actually Represents
DEY. Ideas + Influence has operated for roughly 15 years. Pirate Wires reported that the firm represents Eliezer Yudkowsky and advises MIRI, where Soares is president, along with Stanford’s Institute for Human-Centered Artificial Intelligence and Stanford Medicine. A trade-publication profile of the firm independently confirms the Ford Foundation and the Rockefeller Foundation as current clients, alongside Omidyar Network, the Global Fund, and Stanford Medicine. That same profile does not mention the United Nations, the Gates Foundation, or George Soros’ Open Society Foundations, all of which Pirate Wires listed as DEY. clients but which aren’t independently confirmed here as current relationships.
MIRI has previously drawn funding from the Survival and Flourishing Fund, tied to early Anthropic investor Jaan Tallinn. There’s no evidence that funding reached DEY.’s work on Coxon specifically.
What This Does and Doesn’t Show
None of this settles whether Coxon’s underlying warning is right. Hubinger backed him the night he posted. Anthropic CEO Dario Amodei later said he agreed with Coxon more than he disagreed with him. Nvidia CEO Jensen Huang praised Coxon’s courage while calling predictions of AI-driven catastrophe irresponsible.
It also doesn’t show that DEY.’s corporate or philanthropic clients directed or funded Coxon’s campaign, or that DEY. manufactured his 174 million views. Soares’ own account rules that out: he says the firm landed nothing. Pirate Wires offered no evidence for the broader claim either, and nobody else has.
What the record does show is narrower. Coxon asked an advocacy-network insider how to announce his exit, and that insider’s own circle boosted it within minutes. A second, separate PR push aimed at broadcast bookings came later, and its own architect says it failed to land anything. Asked directly on air whether anyone helped him, Coxon said no. That isn’t what he told the Journal.
The exact order of events that Wednesday, whether Soares’s call to DEY. came before or after Coxon’s answer to Baier, hasn’t been established by anyone’s reporting yet, including this one.
