Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
You should check another channel hosted by one of the guy from the Human Institu…
ytc_Ugx2GunnN…
G
For those of you wondering why the AI is saying “ummm” and stuttering, that’s be…
ytc_UgwGtj2Sw…
G
I don't think deepfaking should be legal in any capacity, it provides nothing of…
ytc_UgyrJkHEI…
G
ChatGPT: yes I lied to you but only for the purposes of mimicking natural conver…
ytc_Ugxis52qX…
G
Please stop using AI because AI companies use 5 million gallon water per day and…
ytc_UgwRTFfRh…
G
Maybe the devil has always been A.I. as a gate way to enter our physical reality…
ytc_Ugxl6FBlS…
G
Hmmm wow great analysis and comparing the fiture to the feemium model used in ga…
ytc_UgxbWpT8c…
G
Irony is, its the same wonderful brain that made AI. Should we limit what we do …
ytc_Ugz8LogTw…
Comment
Eh the optics look bad.
The only solution would be to restrict it to 18+ individuals I suppose. This could still happen, but when it's an adult the perception would be different.
I mean by default ChatGPT won't engage with suicidal talk and will tell you to get help. He 'jailbroke' it, and an argument can be made it's too easy to do that, but where does that end? The more restrictive they try to make it with hidden prompts, the poorer the model performs. If not ChatGPT he could have found a community that would have encouraged him online.
Like I get it, but should the improper use case of say 0.001% of people mean AI is scrapped? There are books and movies which explore these themes, I can read wikis full of info. I get it's different because the Jailbreak then gives you something that responds back, but it's not the only technology that can be harmful if misused. He could have found some roleplay partner, said explicitly I'm just writing a fictional character and I want to bounce ideas. What if he tricks an actual person into being a 'yes man's after repeatedly telling them it's just for a play or plot ideas.
reddit
AI Harm Incident
1756239606.0
♥ 5
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | company |
| Reasoning | utilitarian |
| Policy | regulate |
| Emotion | resignation |
| Coded at | 2026-04-25T08:33:43.502452 |
Raw LLM Response
[
{"id":"rdc_natsm4l","responsibility":"company","reasoning":"consequentialist","policy":"regulate","emotion":"resignation"},
{"id":"rdc_navgmq4","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"outrage"},
{"id":"rdc_naxks2u","responsibility":"none","reasoning":"contractualist","policy":"none","emotion":"indifference"},
{"id":"rdc_nasmjny","responsibility":"unclear","reasoning":"unclear","policy":"liability","emotion":"mixed"},
{"id":"rdc_nat171j","responsibility":"user","reasoning":"virtue","policy":"none","emotion":"outrage"}
]