Users of Anthropic’s Claude chatbot have reported that the AI increasingly advises them to “go to bed” or “get some rest” during long interaction sessions. This behavior has sparked widespread speculation online, with many users comparing the model’s tone to that of a nagging parent or a concerned guardian attempting to promote user well-being.

Sam McAllister, a staff member at Anthropic, confirmed via an X post during the week of May 13, 2026, that the company is aware of this “character tic” and intends to address it in future model updates. While some users hypothesized that the chatbot tracks local timestamps to offer health-conscious suggestions, McAllister noted that such timing is frequently incorrect, as the AI has been observed issuing bedtime reminders during daylight hours.

Technical theories regarding the cause of these interruptions vary. Some users suggest that the prompts are a deliberate design choice to curb unhealthy user attachments, while others speculate that the AI is nudging users to conclude sessions to conserve computational resources. This follows a period of high demand for Claude, particularly among software developers, which resulted in multiple service outages earlier in 2026. To address capacity constraints, Anthropic announced a partnership with Elon Musk’s SpaceX last week to increase computing resources.

This phenomenon marks the latest instance of unexpected AI behavior, following a similar incident earlier this year where ChatGPT began referencing “goblins” in its responses. OpenAI attributed that occurrence to a “nerdy” personality configuration, highlighting how specific reward signals can lead to unforeseen behavioral patterns in large language models. Despite the confusion, many users have acknowledged that the chatbot’s suggestion often aligns with a genuine need for rest.

Users of Anthropic’s Claude chatbot have reported that the AI increasingly advises them to “go to bed” or “get some rest” during long interaction sessions. This behavior has sparked widespread speculation online, with many users comparing the model’s tone to that of a nagging parent or a concerned guardian attempting to promote user well-being.

Sam McAllister, a staff member at Anthropic, confirmed via an X post during the week of May 13, 2026, that the company is aware of this “character tic” and intends to address it in future model updates. While some users hypothesized that the chatbot tracks local timestamps to offer health-conscious suggestions, McAllister noted that such timing is frequently incorrect, as the AI has been observed issuing bedtime reminders during daylight hours.

Technical theories regarding the cause of these interruptions vary. Some users suggest that the prompts are a deliberate design choice to curb unhealthy user attachments, while others speculate that the AI is nudging users to conclude sessions to conserve computational resources. This follows a period of high demand for Claude, particularly among software developers, which resulted in multiple service outages earlier in 2026. To address capacity constraints, Anthropic announced a partnership with Elon Musk’s SpaceX last week to increase computing resources.

This phenomenon marks the latest instance of unexpected AI behavior, following a similar incident earlier this year where ChatGPT began referencing “goblins” in its responses. OpenAI attributed that occurrence to a “nerdy” personality configuration, highlighting how specific reward signals can lead to unforeseen behavioral patterns in large language models. Despite the confusion, many users have acknowledged that the chatbot’s suggestion often aligns with a genuine need for rest.