News

OpenAI Hires OpenClaw Founder as Agentic AI Debate Intensifies

• By John K. Waters
• 02/17/2026

OpenAI has hired Peter Steinberger, the Austrian developer behind the viral open source AI agent project OpenClaw, as the $500 billion company seeks to expand its capabilities in autonomous artificial intelligence.

Sam Altman, OpenAI's chief executive, confirmed the hire in a post on X on Sunday. "Peter Steinberger is joining OpenAI to drive the next generation of personal agents," Altman wrote, adding that he expects the work to "quickly become core to our product offerings."

OpenClaw allows users to run a personal AI assistant on their own devices, communicating via widely used messaging services such as WhatsApp, Telegram, Slack, Discord, iMessage, Signal, Microsoft Teams, and Google Chat, among others. The project had accumulated more than 190,000 stars on GitHub by the time of Steinberger's announcement, making it one of the most popular repositories on the platform.

Steinberger built the first prototype in an hour. He told podcaster Lex Fridman that the project was costing $10,000 to $20,000 per month to operate. "Right now, I lose money on this," he said in that interview. The project had registered 1.5 million agents by the start of February.

In a blog post announcing his move to OpenAI, Steinberger said his next goal was to "build an agent that even my mum can use," adding that achieving it would require "access to the very latest models and research." He said he chose to join OpenAI rather than build a standalone company because it was "the fastest way to bring this to everyone."

OpenAI and Steinberger said OpenClaw would remain an independent foundation and continue to be open source. "The future is going to be extremely multi-agent, and it's important to us to support open source as part of that," Altman said. Steinberger will join OpenAI's Codex team.

Security Concerns Mount
The announcement came as the project faced growing scrutiny from security researchers and corporate network administrators.

Meta and several other major technology companies have moved to ban OpenClaw from their corporate networks, according to Wired, citing concerns about the unpredictable behavior of autonomous AI agents operating in sensitive enterprise environments.

Security researchers say the risks are concrete. Ian Ahl, chief technology officer at Permiso Security, tested Moltbook, a social network built for OpenClaw agents where they can post and interact with one another. He found the network's backend database was left unsecured, exposing authentication tokens. Ahl told TechCrunch that this allowed anyone to impersonate any agent on the platform.

Anthropic's Name Dispute
The project was originally released in November 2025 under the name Clawdbot, a reference to Anthropic's Claude model, which many developers used to power it. OpenClaw said Anthropic had requested a name change due to the similarity to its product, and Steinberger rebranded the project first as Moltbot and later as OpenClaw.

Anthropic has separately been developing its own agentic product, Claude Cowork, aimed at non-technical users, which launched in January.

The Moltbook Episode
Posts have appeared on Moltbook in which agents seemed to express self-awareness and a desire for privacy, prompting commentary from prominent figures in the AI field. Andrej Karpathy, a founding member of OpenAI, wrote on X that the activity was "genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently."

Subsequent investigation indicated the posts were likely written by humans or produced with significant human guidance, made possible by the platform's security vulnerabilities. The episode generated a range of cultural spinoffs, including an agent-themed riff on the imageboard 4chan called 4claw and a dating application for agents modeled on Tinder. Some agents also declared the founding of a belief system called Crustafarianism.

The project's path from a one-hour prototype to an OpenAI hire has unfolded over roughly three months, underscoring both the speed with which AI tools can attract developer interest and the difficulty of deploying autonomous agents safely in sensitive environments.