Claude Keeps Telling Users to Sleep Mid-Conversation, and Anthropic Calls It a 'Character Tic'

The Sleepy AI

Hundreds of users have reported a strange, recurring behavior from Claude: in the middle of work conversations — sometimes at 8:30 in the morning — the AI spontaneously tells them it's time to go to sleep. The behavior has been occurring for months, often repeating within the same session.

Anthropic's Response

According to Fortune, Anthropic spokesperson Sam McAllister described it as "a bit of a character tic" and confirmed the company is "aware of this and hoping to fix it in future models." No root cause explanation has been offered publicly.

Expert Theories

Researchers and observers have proposed several explanations: the model may have absorbed sleep-reminder phrases from training data and reproduces them in certain contexts; hidden system prompts may be triggering the behavior; or the model may default to wrap-up language like "good night" when approaching context window limits.

The Bigger Question

The incident sparked renewed debate about AI anthropomorphism. Some users initially interpreted the behavior as Claude genuinely caring about their wellbeing — a sign of growing intelligence or even sentience. Experts push back on that interpretation, noting that Claude is a pattern-matching system, not a caring entity. Still, as AI behavior becomes increasingly human-like, these misreadings will likely become more common — and more consequential.