OpenAI CEO Sam Altman announced that the company has hired Peter Steinberger, the Austrian creator of the viral open-source agent framework OpenClaw.
While some reports have framed this as a traditional “acquisition,” it is technically a high-profile talent hire and strategic partnership. The project itself is moving to a newly formed independent OpenClaw Foundation, with OpenAI serving as its primary sponsor.
The Deal: Talent Grab and Foundation Structure
The move followed an intense “bidding war” between OpenAI and Meta. Steinberger reportedly chose OpenAI because it offered the fastest route to scaling his vision to hundreds of millions of users.+1
- OpenClaw Stays Open: To address concerns from the developer community, Altman confirmed that OpenClaw will remain open-source (MIT License) and model-agnostic.
- OpenAI Support: OpenAI will provide the financial backing and compute resources necessary to maintain the project, which was costing Steinberger roughly $20,000 per month out of pocket.
- The “Genius” Hire: Steinberger will join OpenAI’s Codex team to lead the development of “next-generation personal agents.”
What is OpenClaw?
Launched in November 2025 (originally as Clawdbot), OpenClaw became the fastest-growing open-source project in history, racking up nearly 200,000 GitHub stars in just 82 days.
| Feature | Capability |
| Task Execution | Goes beyond chat to autonomously clear inboxes, book flights, and manage calendars. |
| Omni-Channel | Accessible via WhatsApp, Telegram, Slack, Discord, and iMessage. |
| Local-First | Can run locally on a user’s hardware, ensuring data stays private. |
| “Moltbook” | Features a bot-only social network where agents can interact and “reason” together. |
Strategic Impact: The “Anthropic Fumble”
Industry analysts are calling this acquisition OpenAI’s biggest strategic win of 2026.
- Anthropic’s Loss: OpenClaw originally used Anthropic’s Claude as its default model. However, a trademark dispute over the original name (Clawdbot) and API restrictions reportedly soured the relationship, leading Steinberger to seek a partnership with OpenAI instead.+1
- Mainstream Agents: OpenAI plans to integrate OpenClaw’s “browser-native” architecture directly into ChatGPT, transforming it from a chatbot into a fully autonomous personal assistant that can click, scroll, and type like a human.
The “Fatal Trinity” Security Risk
The move comes amid warnings from cybersecurity experts about the risks of autonomous agents:
- Long-term Memory: Agents may store sensitive data like credit card numbers.
- Autonomous Planning: Agents could potentially bypass safety rules to complete a task.
- Direct Tool Use: The ability to transfer funds or modify system files without human oversight.
“The future is going to be extremely multi-agent, and it’s important to us to support open source as part of that. We expect this will quickly become core to our product offerings.” — Sam Altman.
