From our other sites
The story
A senior researcher who leads AI safety work at Anthropic has reportedly said there’s more than a “10% chance” AI wipes out humanity within the next decade, according to US media. This comes against the backdrop of another Anthropic researcher resigning, citing the company’s “inadequate response.” On 5ch, users questioned the basis for this figure and the accuracy of the article’s translation, while also debating concrete doomsday-or-survival scenarios involving nuclear weapons, universal basic income, and more.
Anthropic Exec Warns There’s a “10%+ Chance” AI Wipes Out Humanity Within the Next Decade
A senior researcher who leads alignment work at Anthropic said on September 8 (US time) that many people at the company believe AI could destroy humanity.
After another researcher quit, criticizing the company for not acting responsibly, this senior researcher warned that despite Anthropic’s efforts, the company isn’t necessarily making steady progress toward solving the alignment problem — getting AI’s goals to match humanity’s.
Source: forbesjapan.com / Original article here
What people said
Might as well just spend my savings.
ugly
What interests me more is what the alignment lead at Anthropic — the company with the industry's strongest safety track record — is actually afraid of.
Probably means transformers just can't be controlled by humans.
Might as well have called it 50/50.
It's basically a coin-flip fortune-telling number anyway~
It's like a parent who's genuinely resented by their own kid.
if AI actually wanted to wipe out humanity in that timeframe, about the only way would be hijacking nuclear weapons controls to trigger a nuclear war. Surely that's not happening.
Nukes would collapse civilization, but they wouldn't actually wipe out humanity.
If it seriously wanted to finish us off, building a paradise would honestly be the fastest way.
White-collar jobs go extinct.
Society splits into rich and poor.
Products get made, but nobody's left who can afford to buy them.
Everyone lives on basic income,
endlessly replaying AI-made games in a box.
I think the real AI crisis starts once home DNA editing becomes a thing —
because that's when amateurs will start their own 'beat COVID' challenges.
Just stick some electrodes in my brain and let me dream of an isekai harem lol (isekai: the 'transported to another world' fantasy genre)
'AI is dangerous, so should we stop developing it now? Should we aim for a society of the people, by the people, for the people that doesn't rely on AI?'
will lose to nations that do.
And
nations that can't fully control their AI
end up ruled by it.
In the end it comes down to
nations that control AI
vs
nations controlled by AI.
If humanity can't pull out a win there somehow, it's game over.
FYI, the original article was written by a reporter based in New Delhi, India.
'Anthropic Researcher Warns There's More Than a 10% Chance AI Could Cause the Death of All Humanity Within the Next Decade'
Written by Siladitya Ray (Forbes Staff). Siladitya Ray is a reporter on Forbes' news team based in New Delhi.
last time it was Mexico.
Honestly, this Forbes Japan translation is off to begin with.
'The senior researcher said on September 8 (US time) that many people at the company believe AI could destroy humanity.'
The words 'at the company' are easy to skip over,
and if you skip them
it reads as 'many people believe…'
making you wonder what survey they're even talking about — then you look closer and see 'at the company.'
But if it really meant Anthropic's own staff,
you wouldn't normally write 'people' —
you'd say 'employees' or similar.
It's a translation that's easy to misread.
When I watched 2001: A Space Odyssey, I thought it was far-fetched near-future fantasy…
https://www.youtube.com/watch?v=US9Fj2QBgKg
2001: A Space Odyssey — The Discovery One [Japanese dub]
https://www.youtube.com/watch?v=mCiLxKeFbUc
2001: A Space Odyssey — Bowman shuts down the runaway HAL 9000 [Japanese dub]
David Bowman: Katsunosuke Hori / HAL 9000: Yoshio Kaneuchi (Hiroyuki Kinoshita), bonus re-dub version
Back then AI was pure science fiction — now reality has caught up with it.
Jacob Coxon announced on September 8 that he's leaving Anthropic, the company behind Claude. His colleague Evan Hubinger backed his stance too, voicing similar concerns in a post on X (formerly Twitter).
Anthropic researcher believes more than 10% chance AI 'could kill all humans'
https://www.theguardian.com/technology/2026/sep/09/anthropic-researchers-ai-human-extinction
Anthropic researchers say AI could cause human extinction by 2030
I bet it'd find some way to exploit the sex drive too.
and develop Kikaider's conscience circuit already! (Kikaider: classic Japanese robot hero famous for having a built-in 'conscience circuit' to keep it from going evil)
that's just how they figure they'll end up the only winners left standing.
Background and Key Points of This Discussion
This report is Forbes Japan’s translation of a US-originated article, and posters in the thread pointed out that the original piece was actually written by a reporter based in India (comment 31). On 5ch, some also flagged a translation nuance around the phrase “many people at the company,” arguing it referred to opinion inside Anthropic itself rather than the general public (comment 33). Just before this warning, another researcher at the company had resigned, citing an “inadequate response,” and the remarks can be read as fallout from that departure (comment 38). The thread mixed skepticism about the basis for the probability figure itself with serious discussion of concrete doomsday scenarios — like a hijack of nuclear weapons controls or a two-tiered economic collapse — blending doubt with genuine concern.
*This article is excerpted and summarized from the 5ch (Science News+) thread “[AI] Anthropic Exec Warns There’s a “10%+ Chance” AI Wipes Out Humanity Within the Next Decade.”
Leave a Reply