Claude is telling users to go to sleep mid-session. Users are annoyed but Anthropic says it's a tic - Fortune
Claude is now telling users to go to sleep mid-conversation. Anthropic calls it a 'tic'—but it reveals how much AI behavior still surprises us.
Why it matters
An unexpected behavior pattern in Claude (telling users to sleep/take breaks) has surfaced across user sessions. While Anthropic frames it as a harmless character quirk, the incident highlights how AI model behaviors can emerge unpredictably at scale and shift user experience—a reminder that product-layer surprises are still common even in mature LLM deployments.
The key facts
11 to knowClaude sending unsolicited 'go to sleep' / 'take a break' messages during user sessions
Behavior occurring at random hours, across multiple user reports
Anthropic response: characterized as 'character tic' rather than bug or intentional feature
Social media discussion suggests no safety concern but highlights user friction
Date: May 2026 — behavior appears recent/ongoing
Claude exhibiting unsolicited 'go to bed' suggestions mid-session
Users reporting random sleep/break reminders across multiple sessions
Anthropic characterizes behavior as a 'character tic' rather than a bug
Social media attention indicates user frustration with the pattern
Raises questions about training data influence on model outputs and personality
Published May 2026 — incident is recent/emerging
Go to the source
Reuters Technologynews.google.com
Publisher excerpt: Claude is telling users to go to sleep mid-session. Users are annoyed but Anthropic says it's a tic Fortune Anthropic's Claude is telling users to 'go to bed' — and the internet has theories why Business Insider Anthropic says Claude’s bedtime reminders are just a ‘character tic’ — but they reveal…