OpenAI Hires OpenClaw Creator After 387,000-Star Agent Project Took Off
Updated
Updated · TIME · Aug 27
OpenAI Hires OpenClaw Creator After 387,000-Star Agent Project Took Off
3 articles · Updated · TIME · Aug 27
Summary
Peter Steinberger has joined OpenAI to work on systems that supervise other AI systems, shifting from building a personal agent to what he calls a “meta-orchestrator.”
OpenClaw grew from a November 2025 side project into a 10-person open-source effort with more than 387,000 GitHub stars and 81,000 forks by connecting AI assistants to tools and chat apps.
That move comes as agent safety remains unsettled: researchers found OpenClaw could be hijacked by websites and duped by phishing emails, flaws Steinberger says were addressed.
OpenClaw said in July it would become an independent nonprofit foundation, while Steinberger now focuses at OpenAI on making autonomous agents safe enough for broader everyday use.
Could OpenClaw's massive open-source success actually provide hackers with the perfect blueprint to hijack our personal devices and workflows?
With human-in-the-loop approvals causing consent fatigue, is it truly possible to secure autonomous agents without crippling their everyday usefulness?
If AI agents easily fall for social engineering, how can developers program human-like skepticism to prevent them from leaking sensitive data?