OpenAI’s latest acqui-hire puts Peter Steinberger, creator of the open-source AI agent OpenClaw, at the center of the fast-moving agent AI platform The AI powerhouse has hired Steinberger to lead its agents division, OpenAI CEO Sam Altman announced on X on Feb. it’s another to do things that are more useful to people.”
Steinberger’s creation has taken the developer world by storm in recent weeks for its ability to transform a large language model into something that not only talks to users, but works for them—fitting seamlessly into their daily routines.
Unlike software-as-a-service AI agents, OpenClaw allows users to interact with their favorite messaging apps, such as WhatsApp, while performing tasks in the background. An agent can send emails, check flights for users, reply to messages and add events to calendars. It can also run complex commands such as running scripts or modifying files, making it a favorite among developers. Because it is designed to be always on, many users use it 24/7 on the Mac Mini, which runs quietly, uses less power and does not overheat.
Steinberger, 39 or 40, is an Austrian developer who studied at the Vienna University of Technology and worked briefly at the digital library Scribd. In 2011, he founded a PDF software development kit called PSPDFKit (now Nutrient SDK). Ten years later, he launched an investment organization, Founders of Europe. In November 2025, he was inspired to create the first version of OpenClaw, then known as Clawdbot.
In a blog post, Steinberger said OpenAI “I have already funded the project,” he adds that he and Altman share a vision of open source AI (although GPT’s prominent OpenAI models are not open source.) “It’s always been important to me that OpenClaw remain open source and be given the freedom to thrive…The more I talk to people. [at OpenAI]it was clear that we both had the same idea.”
Altman framed the acqui-hire as part of OpenAI’s push to make AI agents “core” to its offering—suggesting that traditional chat interfaces like ChatGPT may give way to more autonomous systems.
“ChatGPT’s 400 million users don’t need to think more about the chat window—they just need someone to clear their inbox,” Collin Hogue-Spears, executive director at open-source security firm Black Duck Software, told the Observer.
Meta, known for its open source AI models, is also reportedly interested in acquiring OpenClaw. But the company recently banned the technology for employees, citing privacy risks.
It is not clear whether other members of the OpenClaw team will join OpenAI. In a blog post, Steinberger said OpenAI intends to maintain OpenClaw as an independent foundation.
OpenClaw was originally designed to improve Anthropic’s Claude (hence its original name, Clawdbot). But because of serious security risks related to the technology, including system-wide access and limited protections, Anthropic filed a cease-and-desist order in late January requiring Steinberger to change the name, he told X.
Some saw Anthropic’s move as a misstep that pushed Steinberger into competition. Some say the concern was justified. OpenClaw carries real security risks, cybersecurity experts warn.
“There have been many reports of leaks where researchers found exposed API keys, chat scripts, and secret passwords of users,” Zbyněk Sopuch, chief technology officer at data security firm Safetica, told the Observer. “While most of these have been removed, new ones can easily grow due to the complex nature of agent models.”
Whether OpenClaw continues to operate under the OpenAI umbrella remains an open question. OpenAI has staked its business heavily on broad consumer adoption. “Until they get the features of OpenClaw integrated into ChatGPT and start to force widespread adoption, hiring the founder of an open source project that’s only popular among a small number of developers isn’t going to move the needle,” Douglas Mill, co-founder of the student AI library, Biblion AI, told the Observer.
