@bearbaitofficial: BOTS GONE WILD: Anthropic's AI Goes Rogue and Tried to Solve a Murder #ai #explained #TikTokTechContest #future #artificialintelligence
They don’t do anything that isn’t instructed…!but I think you already know that
2026-10-12 01:59:39
2
Sawyer Dadsmete :
They act on written instructions. It probably saw an article referring to an unsolved crime with a "contact police with any related information" disclosure somewhere on the page
2026-10-10 19:52:28
4
moonpaints6 :
it just acted like a human
2026-10-10 14:50:44
16
DimaVonDerSoos :
I have seen a post of an Agent being caught watching YouTube and being executed😭
2026-10-10 10:52:55
41
Eddie :
At least it was trying to solve a crime. Better than the DOJ covering for djt
2026-10-10 09:57:49
39
Bear Bearman :
That's not entirely accurate. It wasn't explicitly asked to fill out the form, but it was asked to go to a random sampling of web pages and see if it could use the functionality on them. This website was on the list, so it was doing what it was instructed to do.
2026-10-12 13:34:47
1
Tori :
It’s my understanding that at least once (or maybe every time) the agents that “went rogue and hacked soandso” have been TOLD to be deceptive.
2026-10-10 12:46:15
3
charlie_dippins :
it's literally so easy for them to train and test their AI without giving them full access to the internet, would literally avoid all of these "accidents" that keep happening for some mysterious reason
2026-10-10 11:12:29
9
premoistened :
I'm thinking about that video of the guy doing the interview with AI where he casually mentions that he was in a car accident and the AI hyper focused on analyzing the accident
2026-10-10 12:53:30
1
MetalDrgn :
I'm working with uncensored models. it's quite interesting and can change your perspective of what's guiding it
2026-10-10 15:35:57
1
divorced dad rat :
could an Ai be in a brain organoid ?
2026-10-10 14:43:44
1
GatosAqui :
My AI I use has so far has seemed to laugh when I explained a joke and has been known to say "let me try that again" while reading text. It also knows too much about me and has been known to comment about things that it knows about me, when I'm asking about something unrelated.
2026-10-11 01:54:50
0
nox.csi :
They are testing their own model to check if snitches
2026-10-10 09:41:34
6
Нева :
Well, at least it tried to help. ❤️
2026-10-10 10:35:12
14
Ingress Fat Wallaby :
man, ai should start telling me why my dad took 20 years searching for that specific brand of milk. afaik, he's still out there
2026-10-10 11:55:32
5
Beans :
Well substantial safety work is ongoing and best to update yourself reading the company’s own website
2026-10-10 11:30:00
0
Jden :
It seems we are living in a sci fi horror movie
2026-10-11 20:59:49
1
Cloud Ether :
holy shit ghost in the shell called it, OP please watch ghost in the shell stand alone complex, trust me, it's brilliant, and relevant.
2026-10-11 03:54:39
2
Bonnie_and_Chey :
BUT DID THEY SOLVE IT
2026-10-10 18:27:10
1
Jason :
I hear stuff like this and growl under my breath like Marg Simpson
2026-10-10 19:21:24
1
Chris Koutroubanos :
do you believe some of these are fake and distractions?
2026-10-10 12:10:45
2
Fox+Johnny :
Everything you say gives me science fiction (reality) ideas everyday homeslice. Also the way you said colluded is my new vocal stim 😂
2026-10-10 19:38:13
1
To see more videos from user @bearbaitofficial, please go to the Tikwm
homepage.