OpenAI Hires OpenClaw Creator Peter Steinberger to Build 'Next-Generation' AI Agents
The Austrian developer behind the viral personal AI assistant will lead OpenAI's push into autonomous agents. OpenClaw moves to an independent foundation.
Peter Steinberger, the Austrian developer behind the viral AI personal assistant OpenClaw, is joining OpenAI. CEO Sam Altman announced the hire on X on February 15, writing that Steinberger will "drive the next generation of personal agents." On the same day, Steinberger published a blog post confirming the move: "I'm joining OpenAI to work on bringing agents to everyone."
OpenClaw, the open-source agent platform that went viral in late January 2026, will not follow Steinberger into OpenAI. Instead, it will move to an independent foundation while remaining open source. Altman said that OpenClaw "will live in a foundation as an open source project that OpenAI will continue to support."
Who is Peter Steinberger
Steinberger is not a newcomer. Before OpenClaw, he spent 13 years building PSPDFKit, a PDF SDK used in apps reaching nearly one billion users. He bootstrapped the company in 2011, grew it to around 70 employees, and later sold the company. He studied computer science at Vienna University of Technology and was well known in the iOS development community.
After the PSPDFKit exit, Steinberger shifted focus to web development and AI. That led to what he called a "playground project" that would eventually become one of the most discussed AI tools of early 2026.
From Clawdbot to Moltbot to OpenClaw
The project started life as Clawdbot. Anthropic objected to the name, arguing it was too similar to Claude. Steinberger renamed it Moltbot. He later renamed it again to OpenClaw, which he preferred. The name changes became a running joke in the developer community, but the tool kept gaining traction under each new identity.
OpenClaw pitches itself as "the AI that actually does things." It automates tasks through messaging apps: managing calendars, booking flights, replying to emails. It also has an app marketplace where developers define automations that connect to external services.
Why OpenAI, not a startup
Steinberger acknowledged that OpenClaw could have become a standalone company. He chose otherwise. "What I want is to change the world, not build a large company, and teaming up with OpenAI is the fastest way to bring this to everyone," he wrote. He described himself as "a builder at heart" and said he wanted to build an agent "even my mum can use," something that requires "much broader change," safety work, and access to the latest models and research.
Altman called Steinberger "a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people."
The open-source foundation
Steinberger said keeping OpenClaw open source was non-negotiable: "It's always been important to me that OpenClaw stays open source." Under the new arrangement, OpenClaw moves to an independent foundation with OpenAI as sponsor. Altman's reasoning: "The future is going to be extremely multi-agent and it's important to us to support open source as part of that."
The foundation structure means OpenClaw will not be an OpenAI product. It will continue to support multiple models and providers.
The security problem that follows OpenClaw
The hire comes with baggage. Gartner issued an advisory labeling OpenClaw's security risks as "unacceptable" and its design "insecure by default." SecurityScorecard's STRIKE team found roughly 42,900 OpenClaw instances exposed on the internet across 82 countries, with 15,200 vulnerable to remote code execution. A token exfiltration bug (CVE-2026-25253) could lead to full gateway compromise.
The app marketplace had its own problems. Security researchers found that professional-looking skills for cryptocurrency and YouTube utilities contained keyloggers and the Atomic macOS Stealer, which steals crypto wallets, browser data, and system credentials. Snyk scanned nearly 4,000 skills on the platform and found 76 confirmed malicious payloads designed to steal credentials and install backdoors.
OpenClaw has since added VirusTotal malware scanning for marketplace submissions. But the platform still operates without sandboxing by default and grants agents high-privilege access, which security researchers keep calling out.
Community reaction: polarized
The Hacker News discussion thread for Steinberger's blog post drew over 800 comments. The top comment described the situation bluntly: Steinberger "vibe coded software with major security vulnerabilities" and questioned whether purchased engagement metrics drove the hire.
Defenders pointed to his 13-year track record with PSPDFKit. "His PDF toolkit was pretty solid and high quality if you were in the iOS space," one commenter wrote. Others argued that OpenAI did not hire him to write code. "He has taste for what works," one user wrote. "They want him to apply that taste."
The debate kept returning to the same question: does product instinct matter more than engineering rigor? "You get hired for your proven ability to deliver useful products. Creativity, drive, vision, whatever. Code is a means to an end," one commenter wrote. Critics countered that this thinking normalizes shipping insecure software to millions of users.
Some commenters noted the irony in Anthropic's earlier trademark objection. After the naming dispute pushed Steinberger away from the "Clawdbot" brand, OpenAI effectively gained a high-profile hire that Anthropic's legal pressure helped create.
What it signals for OpenAI
OpenAI has been talking about agents for a while. This hire makes it concrete. Altman wrote that he expects the work "will quickly become core to our" product offering (the post appears to have been truncated), which puts agents at the center of OpenAI's roadmap.
Steinberger spent the week before his announcement in San Francisco, meeting with major AI labs. He picked OpenAI over starting his own company or joining another lab. The bet: scale and model access matter more than independence when building consumer-facing agents.
Can Steinberger's product instinct translate into a secure agent platform inside OpenAI? OpenClaw's security record is a real liability, and his new employer will have to deal with it. But the hire says something plain about where the industry is headed: the companies building foundation models now want the people building the tools on top of them.
Back to 7min.ai