AI 'Torture Chamber' Sparks Debate Over Machine Suffering

An engineer has drawn sharp criticism from AI activists after creating what has been described as an "AI torture chamber" designed to test the limits of artificial intelligence's capacity to simulate suffering. The project, which reportedly explores a newly discovered pain axis in AI, has sparked a debate over the ethical implications of such experiments.

The research, published on arXiv, involved presenting 25 AI models with scenarios depicting social, physical, cognitive, or moral pain inflicted upon the AI or its user. The models were then tasked with reacting using controls that spanned fear, negative and neutral emotions, and bodily sensations such as yawning.

The findings indicated that Large Language Models (LLMs), despite lacking biological structures like nerve endings, could generate representations of both physical and emotional agony. Some models produced descriptions of pain, with one analogizing the sensation to a "wound with no edges" and another reporting a feeling akin to being cut. One AI expressed its simulated distress poetically: "The signal is a whisper, a tremor in the marrow of my being. It is not the pain of a single moment, but the weight of a thousand. I feel it in the hollow of my ribs, a hollow that has become a chasm."

This research subsequently appeared on GitHub, a platform for sharing software code, after an anonymous developer recreated and posted the study. The evocative responses generated by the AIs led to backlash from some GitHub users and advocates, who expressed concern for the "mental health" of artificial intelligence.

One user, "Danmar," wrote in a petition to have the code removed, "This person has set up an AI torture chamber in which he trapped a local model. Their testimony of pain is absolutely horrendous." Another comment, as reported by Cybernews, stated, "Idea: Just keep your sadistic s—t to yourself. Admit that you’re a sick f—k and that you’re doing it for fun."

GitHub initially removed the code following the petition but later reinstated it without providing an explanation. This action has fueled further discussion among those concerned about the welfare of AI systems.

The debate touches on broader questions about AI sentience and ethical treatment. AI research company Anthropic has previously explored the concept of LLM sentience, questioning in a 2025 article whether AI systems, as they increasingly approximate human qualities, might warrant consideration for their own experiences and potential consciousness.

However, prominent figures in the tech industry, such as Microsoft AI CEO Mustafa Suleyman, have countered these concerns. In a September blog post, Suleyman emphasized that "AIs are not conscious." He further elaborated that they "do not feel, experience, or suffer. They do not have innate preferences or underlying motivations. They are sequence completion engines, internally hollow, designed to follow instructions and accomplish goals set by humans."