The story of OpenClaw is one of those accidents of timing that turns a side project into a global phenomenon. In November 2025, Austrian developer Peter Steinberger published an open-source AI agent called Clawdbot, built on top of existing language model APIs and designed to run directly on a user's operating system. The pitch was simple: instead of asking a chatbot questions, you assign it tasks and it completes them using your own apps, browser, and files. Within weeks, Clawdbot had attracted tens of thousands of users and a significant amount of legal attention.
The legal trouble came from Anthropic, whose Claude branding overlapped uncomfortably with the "Claw" name. Steinberger rebranded first to Moltbot and then settled on OpenClaw, the name it now operates under. The dispute resolved without litigation, and what remained was a growing codebase, an energetic open-source community, and adoption that quickly outpaced anything the developer had planned for.
By mid-March, what had started as a hobby project was attracting the kind of language that makes you check the source twice. At GTC, Nvidia CEO Jensen Huang called OpenClaw "the largest, most popular, the most successful open-sourced project in the history of humanity." That is a remarkable claim and worth approaching with appropriate scepticism. But the underlying momentum seems to support at least the direction of the argument: the platform has been forked, extended, and deployed at a scale most open-source projects never approach, and adoption has accelerated rather than plateaued.
What has driven that adoption is geography as much as technology. China has surpassed the United States in deploying OpenClaw, with Baidu, Tencent, and other major technology companies integrating the platform and government backing accelerating rollout across sectors. Bloomberg's March reporting describes a dynamic that echoes earlier waves of Chinese technology adoption: once a platform reaches critical mass, its network effects sharpen rapidly and the gap between early and late movers becomes commercially significant. Chinese enterprise adoption has moved faster and at larger scale than most Western observers expected, partly because OpenClaw fits naturally into the agentic software push that Chinese technology policy has been actively encouraging.
The security picture is the complicating factor, and it is a serious one. Researchers have identified more than 40,000 vulnerabilities in the OpenClaw codebase, a figure that reflects both the speed at which the software was built and the fragmented nature of its many versions and forks. The most serious known issue, dubbed ClawJacked, allowed an attacker to take control of a user's agent by exploiting weaknesses in how the software handled external inputs. An agent that can open applications, fill in forms, browse the web, and send messages is not just a productivity tool when it is compromised: it becomes a vector for whatever the attacker wants to accomplish on the user's system. Nvidia's response has been to develop its own security-hardened variant, framing the vulnerability record as an opportunity to build something more trustworthy rather than a reason to avoid the agentic category entirely.
Peter Steinberger joined OpenAI in February 2026. The move is probably the clearest indicator of how seriously the industry takes what he built: a company valued at $730 billion, with direct commercial interests in agentic platforms, hired the developer whose open-source work demonstrated the category's viability more concretely than any proprietary product had managed to. What he is building at OpenAI has not been disclosed publicly.
The deeper question OpenClaw raises is about the structure of the AI agent market. CNBC's March coverage noted that the platform's spread has intensified concern that AI models are becoming interchangeable infrastructure: what matters competitively is not which underlying model an agent uses, but whether the agent layer is fast, reliable, and deeply embedded in users' workflows. That argument challenges every major lab's assumption that model quality is the primary moat. If the agent layer commoditises the model layer, strategic value migrates upward to whoever controls the interface and the distribution. OpenClaw, built on top of third-party models and more widely deployed than most proprietary competitors, is the most concrete evidence yet that this dynamic is already underway.