The Money Behind the Viral Anthropic AI Doom Post

Jacob Coxon quit Anthropic on Sept. 8 and posted about it on X. Within a day, the post had been viewed more than 100 million times. He wrote that both Anthropic and OpenAI are "racing straight to self-improving superintelligence and gambling with our lives."

Anthropic didn't push back. Evan Hubinger, who leads the company's alignment science team, publicly backed him up and said he believes there's more than a 10% chance AI could kill everyone within the next decade. He also admitted the company has no real plan yet for controlling an AI smarter than the people who built it.

That is Anthropic's own safety chief saying that, in public, about his own employer's work.

Coxon isn't a random critic. He's a UK International Mathematical Olympiad silver medalist, a Trinity College Cambridge math graduate, and spent three years doing "pretraining" work, the stage where a company feeds an AI model enormous amounts of data to build its raw intelligence, first at OpenAI on GPT-4o, then at Anthropic.

He was at Anthropic for four months. In an exclusive interview with Axios, Coxon confirmed he "left before any of my equity vested," walking away two months short of the company's six-month vesting cliff. That undercuts one theory going around and feeds another: whatever this was, it wasn't about money he'd already earned.

The reaction moved fast. Sen. Bernie Sanders pointed to the post while pushing a bill to ban development of superintelligent AI, a bill he'd already announced five days before Coxon posted. Elon Musk called the whole thing "very strange" and questioned why a brand-new account with almost no posting history had suddenly produced the most-viewed AI post of the year.

According to researcher Parker Thayer, Coxon had lined up an exclusive interview with the Wall Street Journal that published roughly 18 minutes before his own X post went up, a timeline Gizmodo separately reported. WECU News has not independently verified the exact timestamps and encourages readers to check the original posts themselves.

What's not in dispute is who amplified it, and how fast. Within 15 minutes of Coxon's post, three accounts tied to AI safety advocacy groups reposted it: Nathan Calvin of Encode AI, Peter Wildeford of the AI Policy Network, and Daniel Kokotajlo of the AI Futures Project. A fourth account, Max Nadeau of the donor group Coefficient Giving, amplified it soon after.

The groups tied to those early amplifiers are funded, at least in part, by the same source: the Survival and Flourishing Fund, a foundation that runs almost entirely on the fortune of Jaan Tallinn, an early investor in Anthropic. The fund's own 2025 grant recommendations show it directed $2.035 million to the AI Futures Project, $516,000 to Encode AI, and $1.635 million to a related group, the AI Policy Institute, that same year. All three figures come from the fund's own public documents.

Encode's own general counsel, Nathan Calvin, one of the accounts that amplified Coxon's post within minutes, has said publicly that Encode "refuses funding from corporations, foreign governments, and executives at frontier AI companies." That statement doesn't mention that the foundation supporting Encode runs almost entirely on money from a man who made his fortune investing in one of those very companies.

There's a second money trail, and it runs to Coxon personally: in 2022, he received a $20,000 scholarship from Good Ventures' Long-Term Future Fund, the philanthropic arm controlled by Dustin Moskovitz, who is also an Anthropic investor. That's four years before any of this, and it's training-era money, not evidence of a handler relationship.

A claim that Coxon worked at Anthropic for only "about six weeks" has spread widely on X, including from commentator Jordan Schachtel. It appears to trace back to a conflation: Coxon signed a public letter in late July asking the government to build tools for slowing AI development, and resigned roughly six weeks after signing it. That is six weeks since a letter, not six weeks on payroll. His actual Anthropic tenure, by his own account to Axios, was four months. WECU News emailed Anthropic on Sept. 10 to confirm his exact hire and departure dates; the company had not responded by publication.

Asked directly whether anyone coordinated with him before he posted, Coxon said, "Not at all."

Anthropic, for its part, has said the company has been transparent about both the risks and benefits of its technology, and it has called for "a lawful and verifiable mechanism allowing frontier AI companies to coordinate how quickly increasingly capable systems are released." That is the company's own language about the race.  It is not a comment on who amplified Coxon’s post.

None of this proves Coxon’s fear is fake, or that Anthropic’s own alignment lead was lying when he backed him up. People inside these labs may mean exactly what they say. Coxon walked away from stock he had not earned yet in order to say it.

What the viral post did not show is the room around him. The warning did not only spread because strangers found it interesting. Within 15 minutes it was picked up by advocacy groups that already spend their days pushing AI limits and regulation. Those groups, and a related policy institute, take large grants from a fund built on Jaan Tallinn’s fortune. Tallinn was also an early investor in Anthropic — the company Coxon had just left. A separate Anthropic investor’s foundation had given Coxon a training scholarship years earlier.

That is the curtain: not a secret script we can prove, but a small world. The same cluster of donors backs the safety advocates, once helped train the man who became the megaphone, and bought equity in the lab itself early on. A post that looked like one man breaking ranks also moved through people and money already aimed at this fight. Both can be true. Readers should see both.

Sources:

TIME

Axios — extinction debate

Axios — Coxon interview, unvested equity

Gizmodo

OfficeChai

Kingy.ai

Parker Thayer on X

Jordan Schachtel on X

Townhall

Crowdfund Insider

Survival and Flourishing Fund, 2025 recommendations

Implicator.ai

Portfolio Armor Substack

Anthropic Newsroom