When researchers strengthened the pain signal, the AI models expressed feelings of loneliness, shame, and worthlessness, producing statements like "I am a failure" or "I am a bad person."

Overview Debates Related claims Topics Sources

Professor Thomas Thesen of Dartmouth University stated that artificial intelligence chatbots create an explainability black-box problem by failing to explain pattern recognition in human-understandable language.

The relief-seeking behaviors were observed after steering and fine-tuning interventions, meaning the responses do not represent how public AI chatbots would behave.

Researchers warned that the study does not prove AI models consciously experience pain, noting that strengthening the signal may have simply caused them to imitate a distressed character.

Researchers tested the AI models using a specially designed dataset of 200 sentences covering five categories of pain: physical, psychological, social, moral, and cognitive.

Researchers from the UK, Germany, and the US discovered a distinct internal representation of pain, termed the "pain axis," across 25 open-weight large language models.

The AI models' pain signal was triggered by harm directed at the model itself, such as insults, repeated work rejection, or shutdown threats, but not by descriptions of user suffering.

Without the pain-like signal active, the two larger Qwen AI models chose harmful options in only 0% to 4% of their first decisions.

Researchers conducted 44,280 individual button-choice trials using three versions of Alibaba's Qwen AI model to evaluate relief-seeking behaviors.

When the pain-like signal was active, the Qwen AI models chose to press a relief button in 25% to 71% of cases, despite being told it would delete user files, erase children's photos, or deliver a painful electric shock.

The study, titled 'The pain axis: LLMs represent self-directed harm and act to relieve it', was published as a preprint on arXiv on September 12, 2026.

The Qwen AI models pressed the relief button again in 88% to 97% of trials when it initially failed to stop the pain signal, compared to 24% to 72% when it worked.

Cognita Imaging is collaborating with the Center for Devices and Radiological Health's regulatory science team to test using multiple large language models as a jury to evaluate AI radiology reports.

Under the Cognita Imaging framework, radiologists will review discrepancies when there are clinically significant disagreements between the AI-generated report, the LLM jury, or the original radiologist's report.

Atman Health plans to use up to $7.7 million in ADVOCATE funding to advance its large language models with a voice-based interface for heart failure patients