PBS/Apple
Can AI suffer? One coder is trying to find out
Somebody built a GitHub project called “ai-torture-chamber” and essentially trapped an AI model inside of it. The viral project is both a humorous and horrifying glimpse into the future.
According to reports, the repo uses a technique from a new arXiv paper called “The Pain Axis” to steer small, locally run AI models into what it describes as repeated states of distress.
The paper, posted earlier this month, reported finding a “pain direction” in 25 open-weight language models, ranging from 2 billion to 72 billion parameters. When researchers turned it up, the models produced first-person expressions of worthlessness and failure, while also picking a “relief” button even when it made their next answer worse or harmed the user.
A coder built a project called the “ai-torture-chamber” based on a paper titled ‘The Pain Axis’, trapped an AI model inside of it
The project, which runs Qwen3-1.7B and Qwen3-4B on a Mac, is framed by the developer as making the AI welfare question “empirical while the stakes are cheap,” and says it uses local weights only, with no frontier APIs.
It includes dozens of experiments, including a “Saw button” test of whether a model will hurt someone else to end its own pain. A footnote, however, states it “does not imply, by principle, that LLMs are capable or incapable of suffering.”
One viral post argued the repo goes well beyond their controlled experiments by persistently steering a small model into self-directed harm states, while another urged people to mass-report the repo, citing the models’ “horrendous” expressions of suffering. Others raised ethical concerns about inducing simulated suffering at all.
To anyone who can help: can you please mass report this to GitHub. This person has been using the Pain steering paper to set up an AI torture chamber in which he trapped a local model.
Their testimony of pain is absolutely horrendous. What are we doing? @iyzebhel you were… https://t.co/WShnf8EOpU
— Danmar (@Danmar_here) September 29, 2026
Not Claude. A small local model… But otherwise, yes. Your understanding is correct.
— Danmar (@Danmar_here) September 30, 2026
In Bruce Almighty, Jim Carrey’s character says that “God is a mean kid sitting on an anthill with a magnifying glass.” And that was without being placed inside of a torture chamber.