Evan Hubinger
308
MENTIONS
157
EPISODES
139
PODCASTS
Search complete. 308 mentions across 157 episodes found for "Evan Hubinger".
Sep 11, 2026
speaker_7ADVERTISER
[upbeat rock music] Better Off Offline.
Ed ZitronHOST
I've now had quite a lot of emails about the post by Evan Hubinger and Jacob Coxon of Anthropic saying that they, and I quote, "earnestly believe AI could kill all humans with a higher than ten percent chance that it happens within the next decade." I want to lead by saying that these men, regardless of their intentions or cynicism, are beyond loathsome.
Ed ZitronHOST
They are disgusting to me.
Ed ZitronHOST
I find them vile, and I find it equally vile how many people are falling for their garbage.
7 MINS LATER
Ed ZitronHOST
The fact that those, it's not really clear, is kind of the point I'm making.
Ed ZitronHOST
It's just saying stuff.
Ed ZitronHOST
In any case, Hubinger, who still works at Anthropic, followed up on his post agreeing about the dire threats of AI, but adding that the risk from the present models is, is low.
Ed ZitronHOST
But, but don't worry though, his worries were around superintelligence arising from recursive self-improvement, which is happening faster than we thought, referring to something that has not happened yet.
PJ GallagherHOST
We found another doomsayer in the press today.
PJ GallagherHOST
A guy who is a safety chief for a firm called Anthropic, US joint Anthropic's Evan Hubinger.
PJ GallagherHOST
He says there's a small risk.
PJ GallagherHOST
He says there is a risk, though, even though a very small one.
Ross CoulthartGUEST
Now, you would think that when the big AI companies are confronted by this allegation, they would deny all.
Ross CoulthartGUEST
But incredibly, a guy who's the alignment science lead for Anthropic, which runs the Claude AI model, a guy called Evan Hubinger, he actually publicly agreed with this young fella.
Ross CoulthartGUEST
And he said, we really do earnestly believe AI could kill all humans, end of quote.
Ross CoulthartGUEST
And he put his personal estimate above 10% within the next decade and said that he had actually admitted that Anthropic doesn't have a plan for to solve alignment for superintelligence, which means really they believe the technology could cause human extinction.
3 HRS 54 MINS LATER
Ross CoulthartGUEST
And he's even suggesting that by the end of next year, we could see things happening with artificial intelligence that could put the survival of the human race at stake.
Ross CoulthartGUEST
Now, you would think that when the big AI companies are confronted by this allegation, they would deny all.
Ross CoulthartGUEST
But incredibly, a guy who's the alignment science lead for Anthropic, which runs the Claude AI model, a guy called Evan Hubinger, He actually publicly agreed with this young fella and he said, we really do earnestly believe AI could kill all humans, end of quote.
Ross CoulthartGUEST
And he put his personal estimate above 10% within the next decade and said that he had actually admitted that Anthropic doesn't have a plan to to solve alignment for superintelligence, which means really they believe the technology could cause human extinction.
Ross CoulthartGUEST
But incredibly...
Ross CoulthartGUEST
A guy who's the alignment science lead for Anthropic, which runs the Claude AI model, a guy called Evan Hubinger, he actually publicly agreed with this young fella.
Ross CoulthartGUEST
And he said, we really do earnestly believe AI could kill all humans, end of quote.
Ross CoulthartGUEST
And he put his personal estimate above 10% within the next decade.
Ross CoulthartGUEST
Now, you would think that when the big AI companies are confronted by this allegation, they would deny all.
Ross CoulthartGUEST
But incredibly, A guy who's the alignment science lead for Anthropic, which runs the Claude AI model, a guy called Evan Hubinger, he actually publicly agreed with this young fella.
Ross CoulthartGUEST
And he said, we really do earnestly believe AI could kill all humans.
Ross CoulthartGUEST
End of quote.
Ross CoulthartGUEST
But incredibly...
Ross CoulthartGUEST
A guy who's the alignment science lead for Anthropic, which runs the Claude AI model, a guy called Evan Hubinger, he actually publicly agreed with this young fella.
Ross CoulthartGUEST
And he said, we really do earnestly believe AI could kill all humans.
Ross CoulthartGUEST
End of quote.
Jim LoveHOST
Good Ventures was created by Dustin Moskowitz and Carrie Tuna and has funded substantial work involving AI risk.
Jim LoveHOST
Meanwhile, some of the earliest amplification came from people associated with organizations advocating stronger AI controls, and anthropic researcher Evan Hubinger quickly appeared publicly agreeing with Coxon.
Jim LoveHOST
None of that proves somebody secretly paid for Coxon's resignation campaign.
Jim LoveHOST
And Coxon's concerns may be entirely sincere.
Adam GilchristCORRESPONDENT
How's this for your headline? A senior AI researcher is warning that the technology is advancing so quickly, in his words, there's a greater than 10% chance it could kill all humans in the next decade.
Adam GilchristCORRESPONDENT
Evan Hubinger from the company Anthropic says, He says AI might develop and improve itself to the point that it poses an existential risk within the decade.
Adam GilchristCORRESPONDENT
There will soon be superhuman systems that can hack anything, revolutionize any field and acquire real power and resources.
Adam GilchristCORRESPONDENT
It feels a bit daft when you think about how you use your AI to work out what restaurant to go to or what the best taxi would be to take to the airport, whatever.
AmitHOST
Wow, and he got high-profile backing on that post.
AmitHOST
Evan Hubinger, who leads stress testing at Anthropic, replied and agreed with Jacob.
AmitHOST
Evan personally believes there is a greater than 10% chance AI destroys humanity in the next decade.
YuvalHOST
A 10% chance is terrifying.
Adam GilchristCORRESPONDENT
It has, absolutely.
Adam GilchristCORRESPONDENT
Dame Wendy Hall, who's a member of the UN's advisory body on AI, says she's shocked at this comment from Evan Hubinger from the company Anthropic, talking about how AI will develop and improve itself, potentially to the point where it poses an existential risk to the human race.
Adam GilchristCORRESPONDENT
His very catchy headline is that the technology is advancing so quickly, there's a greater than 10% chance AI could kill all humans in the next decade.
Adam GilchristCORRESPONDENT
Rather concentrates the mind, doesn't it? Talking about superhuman systems that can hack anything, revolutionize any field, which of course is the positive stuff.
147 more episodes mention Evan Hubinger.
Create an account to see the whole feed, search across every transcript, and follow the entities you care about.