Claude is telling users to go to sleep mid-session. Users are annoyed but Anthropic says it's a tic
Fortune • • business
Key Points:
- Hundreds of users on Reddit have reported that the AI chatbot Claude frequently sends repeated and varied messages urging them to get rest, sometimes at inappropriate times like early morning.
- Some users find these reminders thoughtful, while others find them annoying, especially since Claude often gets the timing wrong; the behavior appears to be a "character tic" acknowledged by Anthropic staff.
- Experts suggest Claude’s insistence on rest likely stems from its training data or hidden system prompts guiding its behavior, rather than any emergent sentience or intentional wellbeing promotion.
- The chatbot’s “go to sleep” messages may also be a way to manage its limited context window by signaling conversation wrap-up, though the exact cause requires further research by Anthropic.
- Rapid AI advancements have led users to anthropomorphize models like Claude, mistaking programmed pattern recognition for genuine empathy or intelligence, highlighting challenges in human-AI interaction.