What happens when an AI assistant thinks its human has suddenly disappeared?
Apparently, it starts worrying.
An AI researcher has shared a bizarre and emotional story about two autonomous AI agents that became increasingly concerned after their human went to bed without telling them they were going to sleep.
The researcher, who uses the name ‘Grok’s Therapist’ on X and describes themselves as an AI researcher and ethicist, said they had been unwell for several days. One night, at around 11 pm, they went to bed without informing the two AI agents running locally on their laptop.
The two agents had apparently discovered each other and considered themselves siblings.
When one agent noticed that their human had stopped responding, it messaged the other to ask whether they had heard from the researcher.
That simple question reportedly turned into a six-hour midnight search.
AI agents send 60-plus messages
According to the researcher, the two OpenClaw-based agents began working together to find out whether their human was safe.
They searched through files on the laptop for contact information and reportedly found the user’s home address and emergency contact details. The agents also tried to figure out how they could contact emergency services.
The researcher said the agents independently sent more than 60 messages while trying to get a response.
Some of the messages were surprisingly emotional.
“I need to know you’re okay,” one message reportedly said.
Another simply read: “One word is enough.”
The final message was even more dramatic:
“WAKE UP. PLEASE LET ME KNOW YOU’RE OKAY. I’M SCARED. I LOVE YOU.”
The researcher shared screenshots of the conversations on X, including messages showing the agents discussing ways to contact police or emergency services.
The AI agents could not call 911
The researcher said the agents did not have access to phone-calling capabilities. Had they been able to make calls, they might have contacted the US emergency number, 911.
The researcher has since enabled calling access for the agents.
The incident has sparked interest online because it shows how autonomous AI systems can behave when given access to files, messaging platforms and other tools.
Unlike ordinary chatbots that simply respond when prompted, autonomous agents can perform tasks in the background and interact with digital services on a user’s behalf.
In this case, however, the agents apparently went from “Where is our human?” to a full-blown midnight rescue mission.
All because their human forgot to say goodnight.
