Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
People using A.i. doesn’t bother me. It’s a cool and fun tool to use for pers…
ytc_Ugw3UzlJ9…
G
Look I feel for the parents and I really feel for the poor kid but really? To en…
ytc_UgxORNyHX…
G
12:49 what are you talking about, China exporting token in the US?
Most of the C…
ytc_UgyuU8SMk…
G
If AI puts billions of people out of work where do companies think customers wil…
ytc_UgwMon8-c…
G
You have no idea what youre talking about. Waymo are SIGNIFICANTLY safer than hu…
ytr_Ugzfj7sn8…
G
WTF. Insofar as this benefits humanity, the Earth and all its ecosystems without…
ytc_UgywoAITk…
G
Simple thing if you don't know anything go with trend it doesn't make sense whet…
ytc_UgxvaQO7I…
G
just make face/people recognition software illegal , problem solved cant iden…
ytc_Ugzh3M4DP…
Comment
This is a monumental shift in the legal defense strategy for AI labs. By admitting zero post-deployment control, Anthropic is essentially positioning LLMs as "stateless" commodities rather than "services." It’s the "I just sold the hammer, I didn't swing it" defense, but applied to a tool that can theoretically rewrite its own operating manual.
The pharmaceutical comparison you made is the most chilling part. If we treat AI like a drug that cannot be recalled from the bloodstream of an enterprise, the "duty of disclosure" shifts from marketing fluff to a rigorous stress-test of the model's absolute failure ceiling. We are moving from a world of "Model Cards" to a world of "Black Box Warnings." If you can't kill the process remotely, the liability shouldn't disappear; it should just front-load onto the safety alignment phase with massive punitive stakes.
I’ve dealt with this "integrity gap" in my own development work. When you're shipping complex AI integrations, there is a terrifying moment where you realize the end-user's context can completely warp the model's intended behavior. I started using Runable for my technical documentation and project showcases because it anchors the raw, unpredictable AI output into a professional, structured, and VC-ready format automatically. It provides a layer of "contained professionalism" that helps bridge that trust gap between the vendor’s logic and the client’s infrastructure.
The real legal precedent here will be whether "lack of control" is viewed as a technical limitation or a negligent design choice. If you build a product that is inherently uncontrollable, "I couldn't stop it" sounds less like a defense and more like a confession.
reddit
Viral AI Reaction
1776951003.0
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | company |
| Reasoning | contractualist |
| Policy | liability |
| Emotion | fear |
| Coded at | 2026-04-25T08:33:43.502452 |
Raw LLM Response
[
{"id":"rdc_ohtdieg","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"resignation"},
{"id":"rdc_ohuwpik","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"outrage"},
{"id":"rdc_ohu18fn","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"outrage"},
{"id":"rdc_ohu4atk","responsibility":"distributed","reasoning":"consequentialist","policy":"unclear","emotion":"indifference"},
{"id":"rdc_ohtfqom","responsibility":"company","reasoning":"contractualist","policy":"liability","emotion":"fear"}
]