Yes, AI Might Really Kill Us All. Plus. . .

The Free Press · collected 2026-09-10 · by The Free Press
Read the original at The Free Press ↗

Summary

A former OpenAI researcher, Daniel Kokotajlo, has written an essay for The Free Press warning that AI might kill all humans due to its rapidly growing power and potential for self-improving superintelligence. This concern is shared by 27-year-old Anthropic and OpenAI researcher Jacob Coxon, who recently announced his resignation citing the companies' prioritization of potentially deadly technology. According to Coxon's post on X, which has been viewed over 125 million times, he believes there is a greater than 10% chance that AI will kill all humans by the end of the decade. Kokotajlo's essay, which is featured in The Free Press today, elaborates on his reasons for leaving OpenAI and supporting Coxon's warning.
Written by the local model on 2026-09-10, using this article's own text rather than the other coverage of the same event (that is the story summary below).

Signals How these are calculated →

Claims extracted
41
claim-shaped sentences
Uncertain
10%
4 of 41 hedged
Leaning
withheld
no quote in the article backed the model's score
Publisher trust
96.2
red-flag proxy, not a credibility rating
Outlets on this story
18
Technology
Narrative spread
1
articles carrying this framing
Analyzed 2026-09-10 · how these are computed

AI analysis (generated at analysis time, not now)

Story summary

Jacob Coxon, a researcher at Anthropic, has resigned from his position after three years working on advanced AI systems. He claims that OpenAI and Anthropic are racing to develop self-improving superintelligence without proper safety measures in place, and that this could lead to humanity's extinction by the end of the decade. Coxon's warning has been viewed over 90 million times on X, and he is not alone in his concerns - hundreds of employees from leading tech firms have urged the US government to support international efforts to manage the pace of AI development. Other researchers, including Evan Hubinger at Anthropic, have also expressed similar worries about the potential risks of advanced AI systems. The controversy has sparked a debate about whether to slow down or "pace" AI development to prevent catastrophic consequences.

Written for “AI poses existential threat” on 2026-09-10, grounded in this article and the 17 other(s) covering the same event.
Why this leaning score
The model judged this article politically coded and scored it +0.35, but every quote it verified points left, so the score is not published.
Written under an earlier scoring contract, which gave a paragraph rather than checkable quotes. Re-analysing this article replaces it.
Leaning score withheld for article 7444: score contradicts its own evidence · logged 2026-09-10

Story

📰 AI poses existential threat
Technology · 18 article(s) covering the same event.

How this is being covered How these are calculated →

Article leaning vs. publisher reliability
Source leaning vs. consistency

Compared with similar articles

This article reads unscored and hedges 10% of its claims. Each row says how that neighbour differs.
The Straits Times
⚖️ Leans left 🔴 25% hedged 2 of 8 📰 publisher trust 94
“Both articles mention Volker Turk warning about AI posing an existential risk to humanity on September 7.”
Daily Mail
⚖️ Leans left 🔴 31% hedged 11 of 36 📰 publisher trust 58
“Both articles mention the warning from Jacob Coxon about AI potentially killing all humans by the end of the decade, and both are dated in September 2026.”
Semafor
⚖️ leaning not scored 🔴 25% hedged 1 of 4 📰 publisher trust 96
“Both articles mention Paul Christiano's warning about potential human extinction due to AI and mention it happening on September 10, 2026.”
We Need a Cure for AI Health Hype different event · 90%
TIME
⚖️ leaning not scored 🔴 7% hedged 2 of 27 📰 publisher trust 95
“The articles cover different topics and do not mention a shared event or incident”
Reason
⚖️ Leans strongly right 🔴 0% hedged 0 of 35 📰 publisher trust 91
“Article A discusses a general topic in tech journalism, while Article B is about the debate over AI and its potential consequences”
BBC News
⚖️ leaning not scored 🔴 27% hedged 7 of 26 📰 publisher trust 96
“Article B is a general commentary on AI risks, mentioning Article A's topic but not describing the same specific happening”
New York Post
⚖️ Leans strongly left 🔴 17% hedged 2 of 12 📰 publisher trust 58
“Article B does not mention the AI researcher quitting or warning about AI's potential harm, instead it appears to be a separate commentary on the topic of AI”
Semafor
⚖️ Leans strongly left 🔴 29% hedged 2 of 7 📰 publisher trust 96
“Article B does not mention AI researchers warning about self-improving superintelligence or the risk of AI killing humans, but rather introduces this topic in a broader context and as an aside”
The Free Press
⚖️ Leans strongly left 🔴 9% hedged 1 of 11 📰 publisher trust 96
“Both articles discuss AI researcher Jacob Coxon's resignation announcement on X, which happened on Tuesday (implied to be September 10, same day as both articles) and are dated closely together”

Publisher

The Free Press · 75 article(s) · 0 correction(s) detected
No corrections detected for this publisher. That may mean careful reporting, or simply that nothing has been checked.

Who wrote this

The Free Press
12 article(s) here · 1 carrying a prediction
🔮 Maya Sulkin talks 9/11, Mamdani, and more with former New York governor George Pataki.
2026-09-10 · assertive framing · Yes, AI Might Really Kill Us All. Plus. . .
🔮 It’s part of an experiment being conducted in a number of places to see if a guaranteed income, for six months or a year, will help keep those newly released from prison from winding up back inside.
🔮 In an interview with our Peter Savodnik days after the shooting, Cox spoke about “which direction he thought Americans, and America, would choose” in the aftermath of the murder.
2026-09-08 · assertive framing · Spencer Cox: The Day Charlie Kirk Died. Plus. . .
🔮 This week, former congressman George Santos was given a lifetime ban from the prediction market Kalshi, after allegedly making $17,000 betting on whether he would go to Donald Trump’s State of the Union address.
2026-09-05 · assertive framing · My Friend Gloria Steinem. Plus. . .
🔮 When the story of the past decade in American public life is written, it will, for better or worse, center on one man.
🔮 On Wednesday, New York City public schools—the largest district in the country—announced that it will ban the use of AI for elementary and middle school students.
2026-09-03 · assertive framing · Should Kids Use AI? A Debate. Plus. . .
🔮 Earlier this week, I couldn’t help but notice a cover story in New York magazine.
2026-08-31 · assertive framing · In Defense of Sports Dads. Plus. . .
🔮 “This might be the last warning we get.” “We are so far past the sci-fi point.” “Full-blown AI takeover within months.”
2026-08-31 · assertive framing · ‘Full-Blown AI Takeover Within Months.’ Plus. . .
🔮 That’s the big question we posed in the latest edition of The Honesty Project, our partnership with Populace, an opinion-research think tank that designs surveys to capture respondents’ “private” opinions—the views they may consider too sensitive to share publicly.
2026-08-27 · assertive framing · Is America Losing Faith in Capitalism? Plus. . .
🔮 How did the daughter of an illiterate sharecropper become one of the world’s foremost lyricists, the woman who wrote more than 3,000 songs, including both “Jolene” and “I Will Always Love You” in the space of a single day?
2026-08-26 · assertive framing · Dolly Parton, Pure Product of America. Plus. . .
Wire or desk byline, not an individual reporter.
More on this subject from The Free Press
AI’s Crisis of Trust. Plus. . .
2026-08-25 · The Free Press · 62% similar
‘Full-Blown AI Takeover Within Months.’ Plus. . .
2026-08-31 · The Free Press · 59% similar
Should Kids Use AI? A Debate. Plus. . .
2026-09-03 · The Free Press · 59% similar
All 12 articles by The Free Press →

Topics

Free Press New York OpenAI The Free Press The Front Page

Subjects

The Free Press ORG · 4× Coxon PERSON · 3× Trump PERSON · 3× Andrew PERSON · 2× OpenAI ORG · 2× Audrey Fahlberg PERSON · 1× Mamdani PERSON · 1× Maya Sulkin PERSON · 1× New York GPE · 1× The Front Page ORG · 1×

Narrative

It was one of the bluntest statements yet of the case for AI catastrophe—and it has coincided with breathtaking new evidence of the technology’s growing power, from last month’s Hugging Face hack to the Millennium Prize math problem Open AI claimed it solved this week.
framing: assertive · carried by 1 article(s) · first seen 2026-09-10
🔮 Maya Sulkin talks 9/11, Mamdani, and more with former New York governor George Pataki.
2026-09-10 · The Free Press
Yes, AI Might Really Kill Us All. Plus. . . · assertive framing

Claims (41 extracted, 4 hedged)

It’s Thursday, September 10. asserted
It → ’ → ?
Audrey Fahlberg reports on Trump’s midterm infomercial. asserted
Fahlberg → report → infomercial
Maya Sulkin talks 9/11, Mamdani, and more with former New York governor George Pataki. asserted
Sulkin → talk → Pataki
Kat Rosenfield reviews Jon Ronson’s new book about the masculinity crisis. asserted
Rosenfield → review → crisis
But first: None of that will matter if the AI kills us all. asserted
AI → matter → us
The stakes of the fast-moving debate over the pros and cons of artificial intelligence are almost too high to fathom. asserted
stakes → move → intelligence
This week, Silicon Valley—and most of the rest of us—are fixated on one unnerving question: Is AI going to kill us all? asserted
AI → fixate → us
It began Tuesday night, when 27-year-old Anthropic and OpenAI researcher Jacob Coxon announced on X that he had resigned. asserted
he → begin → X
The companies, he wrote, “are racing straight to self-improving superintelligence and gambling with our lives. asserted
he → write → lives
The people building AI earnestly believe that it could kill us all by the end of the decade.” uncertain
it → build → decade
By Wednesday night, the post had been viewed more than 125 million times, and counting. asserted
post → view → night
It also set off a wave of responses from AI researchers, including Anthropic Alignment Science Lead Evan Hubinger, who wrote: “[W]e really do earnestly believe AI could kill all humans! uncertain
AI → set → humans
I personally think it is >10% within the next decade. asserted
it → think → decade
The tricky thing about this debate isn’t just that it can get pretty technical pretty quickly. asserted
get → get → debate
It’s that the news is full of doomsday warnings. asserted
news → ’ → warnings
It’s next to impossible to know which to tune out, and which to take seriously. asserted
It → ’ → which
We’ve worked hard to separate the signal from the noise on this issue here at The Free Press. asserted
We → work → Press
And lately, our resident polymath economist Tyler Cowen has injected some healthy skepticism into the conversation about the harm AI may cause. uncertain
AI → inject → harm
But Coxon’s post went viral for a reason. asserted
post → go → reason
It was one of the bluntest statements yet of the case for AI catastrophe—and it has coincided with breathtaking new evidence of the technology’s growing power, from last month’s Hugging Face hack to the Millennium Prize math problem Open AI claimed it solved this week. asserted
it → coincide → problem
So today we’re leading with an essay by someone who, like Coxon, worked at the bleeding edge of AI research before resigning because of what he saw. asserted
he → lead → what
Daniel Kokotajlo is a former OpenAI researcher who resigned in 2024. asserted
who → resign → 2024
Writing in The Free Press today, he explains why he left, and why he believes Coxon’s fears are well-founded. asserted
fears → write → Press
His warning is stark, and worth engaging with: Yes, says Daniel, AI might really kill us all. uncertain
AI → engage → us
A federal judge ruled Wednesday that Andrew and Tristan Tate must remain in custody while they fight extradition to the United Kingdom on rape and human trafficking charges. asserted
they → rule → charges
Andrew faces 42 charges in Britain and Tristan 17, including allegations involving seven victims. asserted
Tristan → face → victims
The Trump administration proposed overhauling the 2030 census Wednesday, seeking to strip undocumented immigrants from the population count used to allocate congressional seats. asserted
administration → propose → seats
The Census Bureau argued illegal immigrants should not be counted because of their “lack of a sufficient tie and allegiance” to the country. asserted
immigrants → argue → country
The proposal would also eliminate questions about race, ethnicity, and sexual orientation from census forms. asserted
proposal → eliminate → forms
The Treasury Department announced it will purchase up to $6 billion in government bonds in an effort to bring down borrowing costs, as the 10-year Treasury yield hit its highest point since 2023. asserted
yield → announce → 2023
Critics say the move fails to address the root cause: national debt that recently topped $40 trillion. asserted
that → say → trillion
Missouri’s congressional maps landed before the Supreme Court for the second time in two days Wednesday, as conflicting state and federal court rulings have left voters uncertain which district they live in ahead of November’s midterms. asserted
they → land → midterms
A resolution is urgently needed, with absentee ballots due to military and overseas voters next week. asserted
ballots → need → voters
President Trump said he came away from a phone call with Vladimir Putin confident that the Russian leader is ready to end the war in Ukraine, after U.S. envoys Jared Kushner and Steve Witkoff visited both Moscow and Kyiv in the past week. asserted
envoys → say → week
“I think more than anything else, it’s two people that hate each other,” Trump said when asked what was standing in the way of a deal. asserted
what → think → deal
The widow of one of five people who died when an Amazon cargo jet slid off a runway at Miami International Airport on Sunday took legal action against Amazon and the operating airline, filing a wrongful death suit yesterday. asserted
jet → die → suit
The suit argues the disaster was avoidable, pointing to pilot error and inadequate training as factors. asserted
disaster → argue → factors
Apple unveiled its first foldable iPhone, called Duo, marking the most significant redesign of the device since its debut in 2007. asserted
Apple → unveil → 2007
The book-style phone, which Apple calls its largest iPhone display ever, will start at $1,999 and hits stores in October. asserted
Apple → call → October
Ben Shelton ousted defending champion Carlos Alzaraz in the U.S. Open quarterfinals early Wednesday morning, completing a five-set comeback that wrapped up at 3:33 a.m. in what became the tournament’s latest-ever finish. asserted
what → oust → 3:33
…and 1 more, not listed.
💬 Give feedback
🕘 History 🎫 Support