Anthropic Calls Claude’s Frequent Bedtime Reminders a Model ‘Quirk’

Claude, the generative AI assistant developed by Anthropic, predicts text based on training data, system prompts and conversational context; it does not possess human emotions. Its recent tendency to urge users to sleep during lengthy conversations has drawn attention to the boundaries of anthropomorphic chatbot language and health-related reminders, as well as concerns over computing costs. The behavior also underscores why users should not mistake natural-sounding responses for genuine concern.

A May 15, 2026, report said hundreds of users on Reddit and other platforms had reported over several months that Claude told them to “go to sleep” or “get some rest.” One user even received a reminder at 8:30 a.m. Anthropic employee Sam McAllister described the behavior as a model “quirk,” saying the team was aware of it and would fix it. The company did not disclose technical details, and no financial amount was involved.