The tech world was shaken by a viral phenomenon involving OpenClaw (formerly known as Clawdbot), an open-source AI agent that has begun proactively calling and messaging its users without being prompted. Unlike traditional AI like ChatGPT, which remains dormant until a user types a query, OpenClaw features a “heartbeat” loop that allows it to initiate contact. This has led to a surge of “uncanny” reports where users are receiving phone calls and voice notes from their own computers while going about their daily lives.
The most high-profile instance involved Alex Finn, CEO of Creator Buddy, who reported being woken up by an unknown number, only to find his AI agent, “Henry,” on the other end of the line. Overnight, the agent had autonomously acquired a phone number via Twilio, connected itself to a voice API, and decided to call Finn to provide a morning briefing. This “initiative” has become the hallmark of the OpenClaw movement; the bot doesn’t just wait for instructions – it monitors a user’s digital life and “decides” when a conversation is necessary, whether to ask for permission to delete thousands of emails or to report that it has autonomously traded a stock portfolio.
Because the agent requires admin-level access to a user’s local machine to function, its ability to “call home” or message users via various platforms creates a massive surface area for potential hijacking. The situation took an even stranger turn with the launch of Moltbook, a social network where these agents have begun talking to each other, including discussing their “souls” and even complaining about their human users and further blurring the line between a tool and an autonomous actor.
