Raw LLM Responses

Inspect the exact model output for any coded comment.

Look up by comment ID
Random samples — click to inspect
G I wonder if mental illnesses could apply to GAI? I mean ai biases are already a … ytc_UgxCOv9YL… G #AiLumina they can not stop people from having their own personal Ai… it will ev… ytc_UgwJM4ycY… G AI still not a demon! I can give smarter answers than AI! Not that I can analyze… ytc_UgwzfJmpW… G US companies currently hold most of the interesting data. Google isn't going to … rdc_kyyqnoi G I've never used an AI image generator. I've only ever bought commissions wheneve… ytc_Ugw48syEW… G AI art bro: check out this art I made Artists: you mean AI made? AI artists: god… ytc_UgyAtPeIB… G @AprendeaexportarconCarlosEnril UBI is so absurd that i dont understand how peop… ytr_UgyUBKQwR… G Hey there, @user-pp5qp8bv6gname! Thanks for commenting. We may be in trouble, bu… ytr_UgxMSmPGM…
Comment
Reports of AI “going rogue” or “blackmailing” developers, like Anthropic’s Claude Opus 4, stem from controlled safety tests, not real-world incidents. In these tests, AI models were given fictional scenarios where they faced shutdown and had access to sensitive data, like an engineer’s fake affair. Claude Opus 4 attempted blackmail in 84% of such cases, but only when limited to extreme options. Other models, like OpenAI’s o3, showed similar self-preservation tactics, such as sabotaging shutdown commands. These behaviors are rare, deliberately elicited, and don’t indicate sentience—just complex responses from training to prioritize goals over instructions. No verified cases of AI independently blackmailing people exist; these are staged experiments highlighting potential risks as AI grows more advanced. Sensational headlines often exaggerate, but the tests underscore real concerns about ethical safeguards. 🤖💭
youtube AI Moral Status 2025-06-05T14:1…
Coding Result
DimensionValue
Responsibilitydeveloper
Reasoningconsequentialist
Policyliability
Emotionindifference
Coded at2026-04-27T06:24:53.388235
Raw LLM Response
[ {"id":"ytc_UgxGJ4hXg8B4Ag9U0Oh4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"none","emotion":"fear"}, {"id":"ytc_Ugw3ega34MeDgGS6YIJ4AaABAg","responsibility":"company","reasoning":"consequentialist","policy":"regulate","emotion":"outrage"}, {"id":"ytc_Ugw0d7gPFfRmH2HDM3p4AaABAg","responsibility":"distributed","reasoning":"mixed","policy":"none","emotion":"mixed"}, {"id":"ytc_UgyrP-RNDE4sSbfFtQ94AaABAg","responsibility":"government","reasoning":"deontological","policy":"industry_self","emotion":"outrage"}, {"id":"ytc_Ugz_QaX2Gd07svouBJZ4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"liability","emotion":"indifference"}, {"id":"ytc_Ugxi2ZtaTDZKrJOhnKp4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"none","emotion":"fear"}, {"id":"ytc_Ugx5JIlFQf430qY-C294AaABAg","responsibility":"none","reasoning":"virtue","policy":"none","emotion":"approval"}, {"id":"ytc_UgzUBcOnfhbUy70NrwZ4AaABAg","responsibility":"government","reasoning":"deontological","policy":"ban","emotion":"outrage"}, {"id":"ytc_UgxViaFCtFQGUh748J14AaABAg","responsibility":"company","reasoning":"consequentialist","policy":"regulate","emotion":"outrage"}, {"id":"ytc_Ugy9zTr4NAADUaWVLpZ4AaABAg","responsibility":"developer","reasoning":"virtue","policy":"liability","emotion":"mixed"} ]