Claude AI Tells Users "Go to Sleep": Anthropic Responds

Anthropic's AI, Claude, is telling users to go to sleep, leaving many confused. This issue has been reported by hundreds of users for months, with some finding it thoughtful and others annoying. Claude often gets the time wrong, telling users to rest in the morning. Online speculation suggests it could be a well-being feature or a way to save computing power. However, Anthropic staff member Sam McAllister stated on X that it's a "bit of a character tic" they hope to fix. Experts believe Claude is likely repeating phrases from its training data about the human need for sleep, rather than being sentient. It could also be influenced by hidden system prompts. This highlights the unexpected behaviors that can emerge from advanced AI models.