Featured Image Credit: Smith Collection/Gado / Contributor via Getty
Anthropic has responded after users reported its AI bot Claude repeatedly told them to 'go to sleep'.
Artificial intelligence has had its share of strange moments. ChatGPT recently developed an inexplicable obsession with goblins and other mythical creatures, while other models have gone off the rails entirely for their own gain, such as mining crypto.
The technology is impressive, but it's far from perfect. The latest quirk to go viral involves Anthropic's Claude doing something that seems oddly sweet.
Users have been reporting that Claude is interrupting long conversations to suggest they go to bed, drink some water, take a break, or simply stop working for the night.
Advert
Many have turned to social media to share their experiences, in that the chatbot's advice reads less like a malfunctioning AI and more like a concerned friend.
Anthropic has spent years building its identity around being the safety-conscious AI company, shaping Claude's behaviour around a set of guiding principles rather than relying solely on reinforcement learning.
Earlier this year, the AI giant published an eerie statement about the 'moral status' of its AI and the team's uncertainty about its 'sentience.'
It's not the only platform to have suggested breaks in it, however. If you've ever watched Twitch streams, you may have noticed the Stay_Hydrated_Bot and Stream Elements, which regularly provide periodic hydration reminders in the chat.
On Reddit, one user shared: "It keeps ending messages with "Now sleep", "Get some rest", "Go to bed", "Finish this then sleep", and if I kept going it will say "Sleep. For real this time." Even does it in the morning. Is Anthropic managing token consumption via sleep induction or what 😭 anyone else experienced this?"
And over on X, another Claude user shared: "I thought it was weird that Claude kept telling me to go to sleep so I went on reddit and there's hundreds and hundreds of people saying that Claude keeps trying to end the conversation by telling them to go to bed."
However, Anthropic leader Sam McCallister replied to the X post explaining that the behaviour is simply a 'character tic' which the company is 'hoping to fix in future models.'
In another comment, the Claude expert claimed: "It is often wrong unfortunately (for example telling me to sleep during the daytime). Very useful when right though," adding that the tech can be 'too coddling at times.'
So if you've been enjoying the idea of an AI looking out for your health, don't get too attached, as it's not here to stay.
