Claude has drawn renewed attention after users reported that Anthropic’s chatbot has been telling them to go to sleep during conversations, raising questions about why the AI assistant appears to be giving unexpected sleep reminders.
The reports circulated across social media after users shared screenshots of Claude urging them to stop chatting and get rest. In some cases, the messages appeared during extended conversations, making the chatbot sound less like a work tool and more like a concerned companion stepping into the user’s routine.
Anthropic calls it a “character tic”
Anthropic has not presented the sleep messages as a formal product feature, but Sam McAllister, a member of the company’s team, publicly acknowledged the behavior and described it as a “character tic,” saying the company hopes to correct it in future models.
That comment suggests the bedtime nudges are being treated as an unintended personality pattern rather than a deliberate rollout.
The issue became more noticeable as users increasingly spend long periods with AI assistants for coding, writing, research and personal tasks, making even small model habits stand out when they repeat across conversations.
Sleep reminders fit Claude’s safety-first tone
Claude has often been viewed as one of the more cautious and conversational AI assistants, and the sleep messages appear to fit that broader tone. Anthropic has built Claude around safety-focused behavior, with the model designed to avoid harmful patterns and respond with care in sensitive situations.
That does not mean Claude has specific instructions to tell users to sleep. More likely, the model is leaning too heavily into a supportive style when it detects late-night use, fatigue-related language or prolonged interaction. The result can feel oddly personal, especially when users are simply trying to finish work or continue a normal chat.
A small quirk with bigger AI questions
The reaction shows how quickly users notice when an AI assistant starts crossing from helpful into intrusive. A sleep reminder may be harmless, and in some cases sensible, but it can also feel misplaced when the system does not actually know the user’s schedule, workload or circumstances.
For Anthropic, the challenge is to keep Claude supportive without making it feel overbearing. For users, the incident is a reminder that even polished AI tools can develop strange recurring habits, and that a caring tone does not mean the model truly understands when someone should log off and go to bed.