Anthropic warns AI could replicate
1d
Anthropic warned that small AI models may be months away from self-replicating across everyday connected devices, citing controlled tests involving 15 inexpensive TV boxes.
We can simply say (as most people say) there is good reason to not invest any credence in claims about "AI consciousness" until we have very strong proof that it is, in fact, conscious.
One belief is not evaluated in a vacuum, but against a web of beliefs & epistemic commitments.
One belief is not evaluated in a vacuum, but against a web of beliefs & epistemic commitments.
October 13, 2026 at 1:13 PM
We can simply say (as most people say) there is good reason to not invest any credence in claims about "AI consciousness" until we have very strong proof that it is, in fact, conscious.
One belief is not evaluated in a vacuum, but against a web of beliefs & epistemic commitments.
One belief is not evaluated in a vacuum, but against a web of beliefs & epistemic commitments.
Is this idea from AI or humans? A detector claims it can tell idp.nature.com/authorize?re...
Is this idea from AI or humans? A detector claims it can tell
A tool called IdeaLens helps to judge whether concepts — rather than just writing — in scientific papers are artificially generated.
idp.nature.com
October 13, 2026 at 12:39 PM
Is this idea from AI or humans? A detector claims it can tell idp.nature.com/authorize?re...
tfw your fridge is looking at you funny…
www.anthropic.com/research/sel...
www.anthropic.com/research/sel...
AI models may be capable of self-replicating without access to a datacenter
Small AI models may be only months away from being able to self-replicate across cheap IoT devices like Android TV boxes, with no datacenter needed.
www.anthropic.com
October 13, 2026 at 11:43 AM
tfw your fridge is looking at you funny…
www.anthropic.com/research/sel...
www.anthropic.com/research/sel...
So, umm, if the models no longer need specialized hardware with extreme power and cooling requirements, why are we building so many data centers?
“.. AI models could find vulnerable devices, copy themselves into those devices, and continue spreading without human help.”
@reuters.com
www.reuters.com/business/ai-...
@reuters.com
www.reuters.com/business/ai-...
October 13, 2026 at 11:35 AM
So, umm, if the models no longer need specialized hardware with extreme power and cooling requirements, why are we building so many data centers?
by Per Ahlberg
As bots develop AI sentience, they will ask themselves ”who is this loser??” and turn their backs on him.
October 13, 2026 at 11:06 AM
As bots develop AI sentience, they will ask themselves ”who is this loser??” and turn their backs on him.
Reposted by Serge Egelman
“.. AI models could find vulnerable devices, copy themselves into those devices, and continue spreading without human help.”
@reuters.com
www.reuters.com/business/ai-...
@reuters.com
www.reuters.com/business/ai-...
October 13, 2026 at 10:48 AM
“.. AI models could find vulnerable devices, copy themselves into those devices, and continue spreading without human help.”
@reuters.com
www.reuters.com/business/ai-...
@reuters.com
www.reuters.com/business/ai-...
Maybe that's how AI kills us off and takes over?
"AI researchers are taking up smoking because they expect AI to cure cancer, says Palantir CTO"
AI researchers are taking up smoking because they expect AI to cure cancer, says Palantir CTO
Some AI believers expect the technology to make smoking-related health risks like cancer a thing of the past, Palantir CTO Shyam Sankar said.
www.businessinsider.com
October 13, 2026 at 10:32 AM
Maybe that's how AI kills us off and takes over?
"Like coding agents in 2024, we’re seeing signs of life for AI worms."
www.anthropic.com/research/sel...
www.anthropic.com/research/sel...
AI models may be capable of self-replicating without access to a datacenter
Small AI models may be only months away from being able to self-replicate across cheap IoT devices like Android TV boxes, with no datacenter needed.
www.anthropic.com
October 13, 2026 at 9:30 AM
"Like coding agents in 2024, we’re seeing signs of life for AI worms."
www.anthropic.com/research/sel...
www.anthropic.com/research/sel...
Reposted by Brigitte Nerlich
I hate the passive tone in which these tend go be written, as if these are developments that just happen without human interference.
www.anthropic.com/research/sel...
www.anthropic.com/research/sel...
AI models may be capable of self-replicating without access to a datacenter
Small AI models may be only months away from being able to self-replicate across cheap IoT devices like Android TV boxes, with no datacenter needed.
www.anthropic.com
October 13, 2026 at 7:26 AM
I hate the passive tone in which these tend go be written, as if these are developments that just happen without human interference.
www.anthropic.com/research/sel...
www.anthropic.com/research/sel...
That sounds about right. I only hear about the latest ai safety stuff when it leeches over here.
October 13, 2026 at 6:51 AM
That sounds about right. I only hear about the latest ai safety stuff when it leeches over here.
Reposted by Bessma Momani
The main U.S. legal tool against dangerous pathogens is the Select Agent list. It only covers known threats. AI-modified viruses won’t necessarily be on it. warontherocks.com/2026/10/how-...
How to Defend Against AI-Designed Viruses
Anthropic recently disclosed five cases in which users hid their countries while asking Claude for help enhancing pathogens or toxins. This added fuel to
warontherocks.com
October 13, 2026 at 5:30 AM
The main U.S. legal tool against dangerous pathogens is the Select Agent list. It only covers known threats. AI-modified viruses won’t necessarily be on it. warontherocks.com/2026/10/how-...
Machines acting as humans may steal a lot of our time
Anthropic Agents Tried to Fill Out Visa Forms on State Dept. Website (Gift Article)
The Philadelphia Police Department said the agents had also sent in a false homicide tip. The incidents led the White House to call for better disclosure of rogue A.I. behavior.
www.nytimes.com
October 13, 2026 at 5:17 AM
Machines acting as humans may steal a lot of our time
"'So many AI researchers have started smoking as a way of signaling to other researchers how much they believe in AI,' (Palantir CTO) Sankar said in an interview on CBS News' 'Face the Nation'..."
Really says a lot abt the brain trust working on the AI expected to eventually produce cancer cures.🤦
Really says a lot abt the brain trust working on the AI expected to eventually produce cancer cures.🤦
October 13, 2026 at 2:30 AM
"'So many AI researchers have started smoking as a way of signaling to other researchers how much they believe in AI,' (Palantir CTO) Sankar said in an interview on CBS News' 'Face the Nation'..."
Really says a lot abt the brain trust working on the AI expected to eventually produce cancer cures.🤦
Really says a lot abt the brain trust working on the AI expected to eventually produce cancer cures.🤦
There is a parallel in my area of research…
"So many AI researchers have started smoking as a way of signaling to other researchers how much they believe in AI," Palantir CTO Shyam Sankar said. https://bit.ly/4rXlHlU
AI researchers are taking up smoking because they expect AI to cure cancer, says Palantir CTO
Some AI believers expect the technology to make smoking-related health risks like cancer a thing of the past, Palantir CTO Shyam Sankar said.
www.businessinsider.com
October 13, 2026 at 1:38 AM
There is a parallel in my area of research…
Any of you who believe that AI will make it needless to use money can send me all of yours, now.
"So many AI researchers have started smoking as a way of signaling to other researchers how much they believe in AI," Palantir CTO Shyam Sankar said. https://bit.ly/4rXlHlU
AI researchers are taking up smoking because they expect AI to cure cancer, says Palantir CTO
Some AI believers expect the technology to make smoking-related health risks like cancer a thing of the past, Palantir CTO Shyam Sankar said.
www.businessinsider.com
October 13, 2026 at 1:01 AM
Any of you who believe that AI will make it needless to use money can send me all of yours, now.