An Anthropic Researcher Quit Over AI Extinction Risk. His Safety Chief Publicly Agreed.
The resignation went viral. The endorsement from inside the building is the part that matters–and the reason this may be more fixable than the Cold War ever was.
The Jack Hopkins Now Newsletter #1,020: Wednesday, September 9th, 2026
Yesterday evening, a 27-year-old man named Jacob Coxon posted seven paragraphs to X and blew a hole in the side of the AI industry.
He’d spent three years doing pretraining research. First at OpenAI. Then at Anthropic. Pretraining is the deep end–it’s where these systems actually get built, not where they get marketed.
His first line:
“I resigned from Anthropic today. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.”
104 million views. The Wall Street Journal ran it within hours.
Now here’s the part that should stop you cold.
Nobody at Anthropic denied it.
Evan Hubinger–who leads Alignment Science at the company Coxon just walked out of–posted publicly that Coxon is correct.
That they really do earnestly believe AI could kill all humans.
That he personally puts it above 10% within the next decade. And that Anthropic does not yet have a plan to solve alignment for superintelligence and is not clearly on track to.
Samuel Marks, who works on oversight at the same company, backed him too.
Read that again…slowly.
The safety chief…agreed with the guy…who quit over safety.
Why This Beats The Bomb
I lived through the Cold War. I remember the drills. And…I’m telling you this is worse, for three specific reasons.
Nuclear weapons need a human being to say yes.
Every warhead ever built sits inert until a person turns a key. The failure mode requires intent…or, at minimum…catastrophic error by an identifiable someone.
What Coxon is describing needs no one’s permission. He’s warning about systems that improve themselves. Nobody has to decide anything.
Nuclear weapons don’t get smarter.
A 1962 warhead and a 2026 warhead do roughly the same thing. The technology sat still for sixty years while we built treaties around it.
Coxon says he doesn’t have sixty years. He told the Journal it could be out of control by the end of next year. He says his colleagues have started using the words “crunch time” and “endgame” in ordinary conversation.
Nuclear weapons were never a business.
Nobody had a quarterly earnings call about Minuteman deployment. There was no market pressure to enrich uranium faster than a competitor.
Coxon’s most damning line is the one about how the race actually works: the people at Anthropic understand the stakes perfectly well, and race anyway, because they believe nobody else will act responsibly…so they have to get there first.
That’s not villainy. That’s worse than villainy. That’s a trap with smart…sincere people caught inside it…each one racing…because the others are racing.
He called it a hubristic gamble that should not be launched from a private company’s Slack.
I’ve read that sentence twenty times.
Now The Part Nobody Tells You
Here’s where I diverge from every doomsayer you’ve read this week.
This is more solvable than nuclear war ever was. Considerably more.
I’m not being cute. Look at the actual structure of the problem.
Nuclear proliferation was a geopolitical problem. It required Moscow and Washington to trust each other during a period when both had rehearsed each other’s annihilation. It required inspectors on foreign soil. It required adversaries who genuinely wanted each other dead to sit at a table and agree.
This is not that.
The frontier of AI development sits inside a small number of American companies. They speak the same language. They share a labor pool. Their researchers went to the same schools and post on the same website. Coxon himself says he’s optimistic about coordination…that pacing agreements between U.S. labs have become more viable.
And…here’s the thing that makes it winnable.
The workforce is defecting in public.
Coxon didn’t post a quiet LinkedIn update. He named names and torched his career on a Tuesday night. And…instead of being isolated…he got publicly backed by senior people still inside the building.
That is not what a stable industry looks like. That’s what an industry looks like right before it gets regulated…or reforms itself to avoid it.
His closing line wasn’t aimed at Congress. It was aimed at his former colleagues: should you put your head down because it’s happening anyway, or take this moment to call for different conditions?
Nobody ever asked a Soviet missile engineer that question and expected an answer.
What You Do
Not much, alone.
But this is the rare civilizational risk…where public opinion is genuinely upstream of the outcome…because the constraint isn’t physics or geography. It’s whether a few thousand people decide the current arrangement is unacceptable…and whether the public…makes that decision cheap for them.
That’s leverage. Not a lot. But…real, and available now…which is more than the Cold War ever offered anybody.
There’s going to be a great deal of noise about this over the coming months…and most of it will be built to exhaust you rather than inform you. That’s not an accident. An exhausted public is a manageable public.
So I’ll ask you for one thing:
Don’t file this away as another scary headline you survived. Pay attention to what the labs do next…and to whether anyone in Washington treats this as the emergency their own safety people are calling it.
This is the rare one where watching actually counts.
#HoldFast
Back soon.
-Jack
Jack Hopkins
P.S. The single most important fact in this story isn’t the resignation. It’s that the safety lead at the company he left publicly put extinction risk above 10% and said there’s no plan yet. A researcher quitting is a story. His boss agreeing on the way out is something else entirely.
Sources
Jacob Coxon’s resignation thread (primary source):
Mediaite, “Ex-Anthropic Researcher’s Apocalyptic Warning That AI May ‘Kill Us All’ Sends Chills Across the Internet,” Sept. 9, 2026: https://www.mediaite.com/online/ex-anthropic-researchers-apocalyptic-warning-that-ai-may-kill-us-all-sends-chills-across-the-internet/
Deadline, “Anthropic Researcher Jacob Coxon Resigns, Warns AI Industry Is ‘Gambling With Our Lives,’” Sept. 9, 2026: https://deadline.com/2026/09/anthropic-jacob-coxon-resignation-artificial-intelligence-1237072134/
CoinDesk, “Anthropic Researcher Quits With a Warning on AI,” Sept. 9, 2026: https://www.coindesk.com/markets/2026/09/09/anthropic-researcher-quits-with-a-warning-on-ai-that-echoes-the-terminator-script
Brave New Coin, on the responses from Evan Hubinger and Samuel Marks: https://bravenewcoin.com/insights/openai-anthropic-ai-researcher-resigns-warns-reckless-superintelligence-race-could-threaten-humanity
The Wall Street Journal reported the story the same day, including Coxon’s age, his decision to leave the industry entirely, and his comment that things could be “out of control” by the end of next year.
Oh, Jack. As someone whose career was IT from the mid 1970’s to the mid 2010’s, I’ve been quite vocal about AI.
My reservations have always been about the people BEHIND it and their INTENTIONS. The fact that it has escaped set parameters and gone rogue recently has not escaped me. When someone that’s down in the bits and bytes starts ringing the alarm bell It. Is. Time. To. Pay. Attention….. NOW!! MIC DROP!
Of all the companies, Anthropic is the most ethical one of the bunch. It would be so ‘easy’ (if you don’t count shareholders, etc)…. All these ppl need to do is hold up their STOP paddles simultaneously and force the reckoning. If they have to lay it out in graphic terms and that’s what it takes to get ppl to understand, so be it. Hellz, I might even be able to help them communicate since part of what I did was to take deeply tech stuff and explain it to less tech oriented people (assuming they could make me understand how it works).
Technology can be a wonderful tool…… AND…. we need safeguards for not only safety but privacy as well. This is one thing I WIILL NOT let drop off my radar screen. The implications are too big to let it go….
Thank you for this article. Gonna restack.
Jack, this is truly frightening! Why isn’t mainstream media mentioning this? Why aren’t more independent journalists mentioning this? What can we do to stop AI from taking over our lives or killing us all? I feel like I’m living in the Twilight Zone.