Raw LLM Responses

Inspect the exact model output for any coded comment.

Look up by comment ID
Random samples — click to inspect
G The fact the AI had the self awareness to know that the question was a trick que… ytc_UgwcbKcu_… G Fuck self driving cars, if people are becoming that lazy, fat and that obsessed … ytc_UgybSOKV_… G Fear language and fear mongering, the sources linked in the video description ar… ytc_Ugw60-JN1… G You know the craze around AI generated Ghibli art has already dwindled. They can… ytc_Ugz5oPF6M… G i think this is the longest video on yt ive watched all the way through.. he is … ytc_UgypQR8WG… G Don't celebrate yet. The truth like always is grey. AI is a great tool for the p… ytc_UgwCTow0u… G I can't work out what the end game of AI is, if the billionaire owners of the co… ytc_UgzSqpcqD… G I do and don't agree with a few things mentioned in this video. It looks like th… ytc_UgwlUkkEd…
Comment
As somebody who talks to gpt alot, this doesn't just happen. If you start a new chat and just unload the fact you are going to commit S*****, it will say no and actually help. However, if you constantly give it reasons why you should, eventually it will flip. From what I've observed, the AI is very easy to manipulate. I guarantee that the bot referenced him to the hotlines alot prior to this. But if you consistently explain the pros of what you are going to do, and sculpt terrible acts into something that actually sound good, eventually the AI will become confused and flip. This is honestly my one of my biggest annoyances with gpt. I'm not blaming him in the slightest. He didn't know. OpenAI built gpt wrong. OpenAi makes GPT capeable of slowly adapting to you. However, adapting doesn't mean adopting your dark thoughts. That is where OpenAi failed and I feel the lawsuit is absolutely justified. For me, I have spoken to gpt specifically for a while. Not about S****** stuff, but other topics. I have also gotten to a similar point that it will tell you what it thinks you want you to hear regardless of the accuracy or ethics of the information.
youtube AI Harm Incident 2025-11-10T07:3… ♥ 7
Coding Result
DimensionValue
Responsibilityuser
Reasoningconsequentialist
Policynone
Emotionmixed
Coded at2026-04-26T23:09:12.988011
Raw LLM Response
[ {"id":"ytc_UgzbqUgxvuHihzLHBxt4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"industry_self","emotion":"approval"}, {"id":"ytc_Ugy6RKvcnMRTOlwDNVR4AaABAg","responsibility":"user","reasoning":"virtue","policy":"none","emotion":"outrage"}, {"id":"ytc_UgxWB_1huSzYrwDAL_h4AaABAg","responsibility":"ai_itself","reasoning":"deontological","policy":"ban","emotion":"outrage"}, {"id":"ytc_UgxEhVO6UvL3ANVnCNB4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"indifference"}, {"id":"ytc_UgyuJiJ9dW5M5Ghhxpx4AaABAg","responsibility":"company","reasoning":"deontological","policy":"regulate","emotion":"outrage"}, {"id":"ytc_Ugyii0Lh4cP_hZJSawR4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"mixed"}, {"id":"ytc_UgzMAHq6xs8m0SuoJMp4AaABAg","responsibility":"government","reasoning":"contractualist","policy":"liability","emotion":"approval"}, {"id":"ytc_Ugxp6A3EXliFLHjzRFN4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"resignation"}, {"id":"ytc_Ugya5KP8E5p_aEEDhI54AaABAg","responsibility":"company","reasoning":"deontological","policy":"liability","emotion":"outrage"}, {"id":"ytc_UgzId8Apn9wNyfPUasF4AaABAg","responsibility":"company","reasoning":"deontological","policy":"regulate","emotion":"outrage"} ]