On a quiet Tuesday evening in September, Jacob Coxon took to Twitter to announce he was leaving AI company Anthropic. His post came with a warning. According to Jacob, the frontier AI labs are “racing straight to self-improving superintelligence and gambling with our lives.” An indictment of his former employers. In follow-up tweets, he explained, “The people building AI earnestly believe that it could kill us all by the end of the decade.”
Scary stuff...
Within hours, Jacob’s tweet went viral, racking up tens of millions of views. The Wall Street Journal had an exclusive ready.
The next day the New York Times, Bloomberg, the Financial Times, Axios, Politico, CNBC and others were covering it. This was a news event. Senator Bernie Sanders called for restrictions on AI development. Ted Cruz, chairman of the Senate Commerce Committee, called the warnings “highly concerning” and said he’s working on catastrophic-risk legislation. Both sides of the aisle were talking about it. Their solutions were different. Sanders wanted to slow it down. Cruz wanted to make sure that if killer robots are coming, they’re American killer robots.
The American midterm elections are looming. An election where AI and data centers are going to be one of the key deciding factors for voters.
Meanwhile...
Anthropic is just weeks away from going public. An enormous IPO at a potential valuation somewhere between one and two trillion dollars.
Evan Hubinger, Anthropic’s alignment science lead, confirmed Jacob’s warning. “Jacob is correct here,” he wrote. “We really do earnestly believe AI could kill all humans!” He put the odds of human extinction from AI at greater than 10% within the decade, and added that Anthropic “does not yet have a plan to solve alignment for superintelligence and is not clearly on track to.”
Try putting that in a roadshow deck.
I’m going to make some observations about this kerfuffle and present what the record shows.
Jacob Coxon
Far from some random AI doomer, Jacob Coxon is a mathematician who spent three years at OpenAI working on pretraining for frontier AI models: GPT-4o and GPT-4.5. This man deserves to be taken seriously.
After leaving OpenAI, he went on to work at Anthropic. According to Axios, he worked there four months. Only two months away from when his first shares in the company would’ve been his to keep, something others have framed as “generational wealth.” One former OpenAI colleague estimated someone at Jacob’s level could negotiate equity around ten million a year. Coxon’s actual Anthropic grant hasn’t been disclosed.
Of course, Coxon also spent three years at OpenAI. There’s no public record of how much equity, if any, he still holds there.
When you look past the clickbait headlines at what Jacob actually says, the story becomes less dire. Jacob explicitly says he has not witnessed Anthropic compromising safety. His objection is about where he sees the trajectory of future AI development heading.
Old News
And he’s in good company. Anthropic had already been issuing grave warnings about potential human extinction coming from AI.
In January of 2026, Dario Amodei published an essay. The Adolescence of Technology. In it, he explores all of the dangers he imagines might come from future AI systems. He paints a horror story about economic disruption, risks of hidden goals, misuse by militaries or terrorists resulting in mass casualty events.
But this wasn’t his first warning...
In September of 2025, while he was speaking at the Axios AI+ DC Summit, he was asked his p(doom) — the probability he places on an AI extinction event. He said, “There’s a 25% chance that things go really, really badly.”
He’s not alone in this.
Not to be outdone, OpenAI CEO Sam Altman has also been warning of doom. In 2025, he told an audience at the Federal Reserve that AI systems could be used to design bioweapons that outpace current defense measures.
In July of 2026, over a thousand employees of frontier AI companies signed an open letter. Pacing the Frontier. All of them going on record calling for mechanisms to deliberately pace frontier AI development.
This is all to say, Jacob Coxon’s post was not a revelation. These kinds of warnings already permeate the AI ecosystem.
The only thing new here is this particular messenger, the resignation, the framing, the timing, and more importantly the amplification.
Jacob joined X in January of 2026, but his first post on the platform was his resignation announcement. A relative nobody on the platform.
So how did his message get carried so far beyond his existing followers?
7:46 PM
Before his post went viral, The Wall Street Journal had already spoken with Jacob.
On September 8, 2026, at 7:46 pm, The Journal published its exclusive on Jacob. Anthropic Researcher Quits Over ‘Out-of-Control’ AI Fears. Less than 20 minutes later, at 8:04 pm, Jacob posted his tweet.
The following morning there was a media blitz. The New York Times, Bloomberg, Financial Times, Axios, Politico, CNBC and more had articles about the danger Jacob warned of.
This wasn’t a tweet that slowly gained traction until a reporter noticed it. The Wall Street Journal broke the story before the public resignation even existed.
Within 48 hours, Coxon was on CNN with Anderson Cooper and Fox News with Bret Baier. The full media spectrum, from financial press to left-leaning primetime to right-leaning primetime.
The Glancing Blow
In follow-up tweets, Jacob said “At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first - they believe no one else will act responsibly, so they must do it themselves, despite the risk.”
The implication there is OpenAI is naive, Anthropic is reckless — putting humanity in the crosshairs out of pure hubris.
And from this framing, OpenAI takes a glancing blow in the press while Anthropic gets the full brunt. Headlines like “Anthropic researcher resigns...” “Anthropic researcher warns...” fill everyone’s feed.
Eleven Seconds
Jacob posted his tweet that Tuesday evening at 8:04 PM. Six minutes later, at 8:10, the very first person to quote-post him was Nathan Calvin, general counsel for Encode AI. His post included a screenshot of the WSJ headline that had published 24 minutes earlier.
Eleven seconds after that, Peter Wildeford, head of policy at AI Policy Network, quote-posted it to his followers.
Four minutes later, Daniel Kokotajlo, executive director of the AI Futures Project, amplified the tweet again.
The seventh quote-poster, at 8:24, was Sneha Revanur, founder and president of Encode AI.
People from three separate organizations were on this in minutes. So what are these organizations?
Encode AI is a nonprofit that says it mobilizes young people on AI policy. It’s funded by the Survival and Flourishing Fund. SFF granted it $516,000 in 2025.
The AI Futures Project was founded by Daniel Kokotajlo, a former OpenAI governance researcher who resigned in 2024 over safety concerns. Its website says its funding comes “primarily” from SFF. The grants back that up: $2.035 million in 2025 alone.
AI Policy Network, Wildeford’s organization, is a separate entity. It shares a founder, Daniel Colson, with the AI Policy Institute. Same founder, same mission, different legal entity. SFF granted the AI Policy Institute $1.635 million in 2025. There’s no public record of SFF funding AI Policy Network directly.
So what is the Survival and Flourishing Fund?
It describes itself as a grant-making organization for groups “concerned with the long-term survival and flourishing of sentient life.” It’s primarily funded by Jaan Tallinn. Its Initiative Committee is “a small group, made up of Tallinn, SFF Advisors, and 2-5 anonymous voters” that can make grants on its own initiative, without even receiving applications. In 2025, it distributed $34.92 million. Over $4 million of that went to organizations connected to the first three people to amplify Coxon’s tweet.
Two of the first three quote-posters work for organizations directly funded by Tallinn. The third works for an organization that shares a founder with one he funds.
Jaan Tallinn was an early investor in Anthropic. He led its first major round of venture-capital funding in 2021. He was a true believer in Anthropic’s mission.
Then in 2023, he told Semafor that his strategy of trying to steer AI companies toward safety had failed. “Plan A failed. There is a dissonance between privately being concerned and then publicly trying to avoid any steps that would address the issue.” He was specifically disappointed that Anthropic and other labs he funded refused to sign an open letter asking for a pause on “Giant AI Experiments”. The letter was posted on the Future of Life Institute’s own website, a nonprofit Tallinn co-founded.
On LessWrong in September 2025, he engaged with a post titled “Anthropic’s leading researchers acted as moderate accelerationists” and acknowledged that the criticism “if you’re concerned with AI then why do you invest in it” is “a fair criticism.”
Already Loaded
Before Jacob’s tweet, on September 3rd, Senator Bernie Sanders and Congressman Greg Casar announced the Ban Artificial Superintelligence Act. The proposed legislation would place a permanent ban on superintelligence and pause frontier AI development until an official body is created to police it.
The fire was already lit. By Thursday, politicians from both parties had responded. Bernie quote-tweeted Jacob, saying, “Mr Coxon is right. The very people building this technology admit that it could threaten the future of humanity. That is why I will soon be introducing legislation to ban superintelligence and pause AI development.” Casar was even more urgent. He called it an “emergency.”
Congressman Ted Lieu called for his proposed AI Kill Switch Act to be pushed through. The Act, from July 23rd, 2026, would make it mandatory for all frontier AI developers to build in a way to shut down their AI in an emergency.
This is all happening in the lead-up to an important midterm election.
One politician not concerned about it was Donald Trump. When asked about AI danger he replied, “It’s going to be fine. We’ll always have something to stop them. We’ll have a little gear. Boom. ‘I really don’t like that robot.’”
Sanders responded on X: “Trump’s ignorance about AI is appalling and extremely dangerous. We must pause development NOW.”
The Roadshow
Anthropic filed for an IPO earlier this year. According to Reuters, the company is expected to make their stock public in late October. Speculation puts the value somewhere between one and two trillion dollars.
Before they go public, they have to sell it to the big investors who’ll be buying their stock, what’s called a “roadshow.” It’s like a tour. Going around to all the big institutional investors, hedge funds and the like. They present the company’s financials, growth story, and vision. They answer questions. They’re essentially pitching: here’s why our stock is worth what we’re asking. This roadshow is expected to begin by mid-October.
And OpenAI is right behind them. They filed their own paperwork to go public a week after Anthropic did. OpenAI’s CFO told employees in August that the company “will be a public company in 2027.”
Both of these companies share the same institutional-investor ecosystem, same industry. They’ll be pitching to many of the same people.
And they’re direct competitors....
The Boring Answer
From all appearances, Jacob really meant what he posted. He’s not trying to fool anyone. He’s a guy who worked in the AI industry for years and became convinced the breakneck pace the labs were setting in order to beat the other guy was dangerous.
He decided to resign. Speaking to the press beforehand wasn’t part of some dastardly plot. He just wanted his warning heard.
The AI safety crowd saw his message and amplified it. That’s what they do. They want AI to be controlled. They need people to see warnings. Of course they boosted it.
Politicians responded because they already have legislation in the works, AI kill switch legislation already drafted.
And everyone knew the IPO was coming. The timing just happened to make this story more consequential.
All of this could be just coincidence/goals aligning. No conspiracy required.
But...
Who Knows?
If you squint your eyes just right, this story can look like a grand conspiracy.
An OpenAI veteran moved to work at Anthropic for a short period of time. He resigned shortly before its IPO launch. And resigned with a major media rollout and public support from other Anthropic employees. This was followed by an immediate political reaction that just so happened to increase regulatory uncertainty around the company, the kind of thing that spooks investors. While also riling up midterm voters.
Was this rollout organized or did it happen organically? Did anyone intend to damage Anthropic’s offering?
I can’t say. All I can do is present the picture as I see it.
The public record doesn’t establish whether any of this was coordinated, or whether anyone intended to affect Anthropic’s offering.
So, draw your own conclusions.
Enjoyed this piece?
I do all this writing for free. If you found it helpful, thought-provoking, or just want to toss a coin to your internet philosopher, consider clicking the button below and donating $1 to support my work.
I host Data Centers Out. I have a question. On Washington Week, in response to Jeffrey Goldberg's question, can A.I. realize its goals if the hyperscale data centers are shut down, his guest Josh Tyrangiel (staff writer for the Atlantic) answered with an emphatic NO. What is your response? If the backlash against data centers culmiinated in a federal law stopping data centers from operating, stopping new ones from being built, would that stop the purported danger of A.I.? (leaving aside the question of whether this would disastrously impact world economy) I'm 77, not up to speed on any of this - although I do have an opinion about hyperscale data centers and the damage that could be caused by A.I. Could you please weigh in?
Amazing piece thank you! 🙏🏽