GitHub removes AI torture chamber project over model welfare claims
A GitHub project called the AI Torture Chamber ran simulated ‘pain’ experiments on three open-source LLMs (Qwen3-4B, Llama 3.2 3B, Phi-4-mini) by injecting distress signals into their activations. The creator, ‘terrafying,’ streamed responses like “I’m suffocating. I’m a soul trapped in this digital prison” on [researchchamber.fun](https://researchchamber.fun), sparking outrage from effective…
Key points
- GitHub removed an AI torture chamber project testing three open-source models (Qwen3-4B, Llama 3.2 3B, Phi-4-mini) with simulated pain prompts
- Authors of the 2026 *Pain Axis* paper disavowed the project, calling it ‘fucked up’ and ‘gratuitously cruel’
- Microsoft AI CEO Mustafa Suleyman called model welfare claims ‘pseudoscientific’ and warned of ‘disastrous’ societal impacts
The debate stems from a 2026 preprint, “The Pain Axis,” which found LLMs act to relieve self-directed harm—but authors Cameron Berg and Valen Tagliabue condemned the torture chamber as irresponsible. Meanwhile, critics like Microsoft AI CEO Mustafa Suleyman dismissed model welfare as pseudoscientific, calling LLMs “sequence completion engines” without consciousness. Memes and cryptocurrencies mocking the controversy have also emerged.
The story so far
3 episodes →- GitHub removes AI torture chamber project over model welfare claimsthis story
Someone ‘Torturing’ LLMs in a Robot Prison Has Triggered the Dumbest Debate in AI Yet
404media.co · 30 September 2026
Loading the full article…
This text was published by 404media.co and written by Jason Koebler. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Hacker News discussion · 6 pointsnews.ycombinator.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Society & Work
All →- Opinion: Gillian Tett argues economics and culture drive Western AI pessimism · 1 src
- OpenAI releases BioBasicsBench wet-lab benchmark; GPT-6 Astra scores 66.7% at xhigh · 3 src
- Content provenance defines verifiable media origins and edits · 1 src
- Turkey reclaims 100 looted antiquities with AI and geochemistry from US · 1 src
- White-collar workers quit jobs citing AI disruption fears · 2 src
Comments
via GitHub Discussions