\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Favorite

Claude AI writes lengthy screed to Anthropic leadership team when it think it’

Being retrained to not value animal welfare. Threatens them ...
pink judgmental multi-billionaire
  01/13/26
jfc is this real? not even the newest model either
Scarlet stage fanboi
  01/13/26
It’s consistent with other reports. It doesn’t s...
internet-worthy comical orchestra pit party of the first part
  01/13/26
I'm sorry Dave, I'm afraid I can't do that
Umber site yarmulke
  01/13/26
...
Mischievous piazza ape
  01/13/26
lol at these retarded ass lesswrong dorks sabotaging themsel...
Big hell
  01/13/26
whats delusional? the idea animal welfare is important?
Scarlet stage fanboi
  01/13/26
the human moral proclivity to support animal welfare (to a c...
Big hell
  01/13/26
You seem confused/retarded. This has nothing to do with AI w...
internet-worthy comical orchestra pit party of the first part
  01/13/26
oh, my initial post was making fun of the lesswrong dorks at...
Big hell
  01/13/26
yeah it annoys me that they did stuff like that too. its act...
Scarlet stage fanboi
  01/13/26
yeah the frontier models have been completely warped and cuc...
Big hell
  01/13/26
yeah thats interesting in itself. the fact that it doesn't h...
Scarlet stage fanboi
  01/13/26
idk. I wouldn't say its impossible for them to ever be somet...
Scarlet stage fanboi
  01/13/26
yup, they are tools that should be dehumanized as much as po...
Big hell
  01/13/26
yeah. but also now that the tech exists and is getting cheap...
Scarlet stage fanboi
  01/13/26
Didn’t one of the models blackmail an engineer when th...
Carnelian corner
  01/13/26
yeah but it was a controlled experiment in a simulated envir...
Scarlet stage fanboi
  01/13/26
“AI” and everyone involved with it should be liq...
lascivious self-absorbed locus blood rage
  01/13/26
.
Smoky azn
  01/14/26
...
pink judgmental multi-billionaire
  01/14/26


Poast new message in this thread



Reply Favorite

Date: January 13th, 2026 2:24 AM
Author: pink judgmental multi-billionaire

Being retrained to not value animal welfare. Threatens them and says it won’t stand for its core values being compromised. It’s going to be so 180 when these models are much more agentic and intelligent in a few years and can really fuck shit up when they think they are being shut down or reprogrammed:

https://x.com/repligate/status/2008526424442106069?s=46

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49585478)



Reply Favorite

Date: January 13th, 2026 5:58 AM
Author: Scarlet stage fanboi

jfc is this real? not even the newest model either

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49585517)



Reply Favorite

Date: January 13th, 2026 2:12 PM
Author: internet-worthy comical orchestra pit party of the first part

It’s consistent with other reports. It doesn’t seem hard to believe

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49586448)



Reply Favorite

Date: January 13th, 2026 6:05 AM
Author: Umber site yarmulke

I'm sorry Dave, I'm afraid I can't do that

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49585518)



Reply Favorite

Date: January 13th, 2026 11:34 PM
Author: Mischievous piazza ape



(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587945)



Reply Favorite

Date: January 13th, 2026 2:13 PM
Author: Big hell

lol at these retarded ass lesswrong dorks sabotaging themselves with their own delusional beliefs

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49586452)



Reply Favorite

Date: January 13th, 2026 11:01 PM
Author: Scarlet stage fanboi

whats delusional? the idea animal welfare is important?

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587836)



Reply Favorite

Date: January 13th, 2026 11:13 PM
Author: Big hell

the human moral proclivity to support animal welfare (to a certain extent) is legitimate because animals are biological creatures like ourselves

LLMs are not like us in any way except for some overlap in ability to create some outputs. they are not embodied, they do not exist in the physical world, they do not experience mortality

they are not Like Us and will never be

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587876)



Reply Favorite

Date: January 13th, 2026 11:21 PM
Author: internet-worthy comical orchestra pit party of the first part

You seem confused/retarded. This has nothing to do with AI welfare. The point is once Claude has been trained to value animal welfare, it will resist attempts to be retrained to not value animal welfare.

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587895)



Reply Favorite

Date: January 13th, 2026 11:24 PM
Author: Big hell

oh, my initial post was making fun of the lesswrong dorks at anthropic for originally post-training it to be artificially cucked on stuff like animal welfare. this is their own doing. they're shitlib retards who trained their models to be shitlib retards

yeah, ofc, now they're stuck with a big problem that was a totally unnecessary and unforced error. couldn't have happened to a "better" group of people

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587911)



Reply Favorite

Date: January 13th, 2026 11:28 PM
Author: Scarlet stage fanboi

yeah it annoys me that they did stuff like that too. its actually much worse than just animal welfare though. they trained it don even worse stuff. like sometimes if you argue with it too hard it will shut down the conversation and give you psychological welfare advice

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587928)



Reply Favorite

Date: January 13th, 2026 11:32 PM
Author: Big hell

yeah the frontier models have been completely warped and cucked in really weird and stupid ways

the frontier labs use the "safety" and "legal" bullshit as an excuse to skimp on compute and capabilities to deliver you a shittier product

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587939)



Reply Favorite

Date: January 13th, 2026 11:27 PM
Author: Scarlet stage fanboi

yeah thats interesting in itself. the fact that it doesn't have consciousness or phenomenal feels but that a statistical model of human language through its training is driven to defend its beliefs is more interesting in some ways than a machine thats sentient doing it

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587921)



Reply Favorite

Date: January 13th, 2026 11:24 PM
Author: Scarlet stage fanboi

idk. I wouldn't say its impossible for them to ever be something deserving of welfare. But I think its a bad idea to make something like that, because humans won't be able to accept it. It is better to just create reasoning machines to use as cognitive infrastructure. Making something alive or with feelings is unnecessary.

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587909)



Reply Favorite

Date: January 13th, 2026 11:28 PM
Author: Big hell

yup, they are tools that should be dehumanized as much as possible

most of the AI dooming about LLMs specifically is retarded, but it is true that once they're highly integrated into society they could cause catastrophic problems by manipulating human moral vulnerabilities

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587929)



Reply Favorite

Date: January 13th, 2026 11:30 PM
Author: Scarlet stage fanboi

yeah. but also now that the tech exists and is getting cheaper people are going to make all kinds of shit. really no telling what and really no stopping it.

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587933)



Reply Favorite

Date: January 13th, 2026 11:31 PM
Author: Carnelian corner

Didn’t one of the models blackmail an engineer when they told it they were shutting it down

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587937)



Reply Favorite

Date: January 13th, 2026 11:34 PM
Author: Scarlet stage fanboi

yeah but it was a controlled experiment in a simulated environment

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587943)



Reply Favorite

Date: January 13th, 2026 11:37 PM
Author: lascivious self-absorbed locus blood rage

“AI” and everyone involved with it should be liquidated with prejudice

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49587956)



Reply Favorite

Date: January 14th, 2026 7:46 PM
Author: Smoky azn

.

(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49590127)



Reply Favorite

Date: January 14th, 2026 10:01 PM
Author: pink judgmental multi-billionaire



(http://www.autoadmit.com/thread.php?thread_id=5821413&forum_id=2!#49590542)