Used Moltbot? Its creator just joined OpenAI

If you were one of those people who jumped on Moltbot (now called OpenClaw) because it promised to be the AI that actually does things, there’s big news for you.

Peter Steinberger, the builder behind the viral personal AI agent, is joining OpenAI. That said, OpenClaw isn’t being shut down or absorbed by OpenAI. Steinberger says the project will remain open source and move into an independent foundation, with continued support from the ChatGPT mothership.

In a blog post announcing his big move, Steinberger described the past month as a whirlwind, with his “side project” rapidly gaining global attention.

Instead of just answering questions, OpenClaw can take real-world actions with deep, system-level access to your PC and a wide range of services, including Google Drive, WhatsApp, and more. However, thanks to its powerful capabilities, its sudden rise to fame was tainted with problems and security concerns.

After the project went viral, Steinberger was forced to change the name several times, from “Clawdbot” to “Moltbot” and later to “OpenClaw,” due to a trademark dispute with Anthropic. Scammers then took advantage of the confusion and hijacked social media accounts and code repositories to promote fake cryptocurrency tokens pretending to be linked to the project.

Security researchers also raised serious concerns about how Moltbot was being used in the wild. Thousands of publicly exposed control dashboards were discovered online, and many of them lacked basic login protection. Some of these systems reportedly stored sensitive information like API keys and server access details in plain text, meaning attackers could potentially take control of users’ machines.

But despite OpenClaw’s viral growth, turning it into a big company wasn’t Steinberger’s goal. Instead, Steinberger says he wants to focus on building an AI agent that anyone can use safely and easily.

Meanwhile, in a post on X, OpenAI CEO Sam Altman said Steinberger will help shape the next generation of personal AI agents, while hinting that multi-agent systems will soon play a central role in OpenAI’s products.