Author
Huang Kaiyue, Research Assistant of CGAIG
Since the beginning of 2026, OpenClaw has rapidly become a focal point in global discussions on the AI industry and policy. The open-source AI agent—released in November 2025 by Austrian developer Peter Steinberger—has accumulated more than 325,000 stars on GitHub within just four months, reportedly making it one of the fastest-growing software projects on the platform. However, although the project originated in Europe, much of its subsequent progress has unfolded largely outside the region.
In February, Steinberger announced that he had joined OpenAI and relocated to San Francisco, while the project itself transitioned to an independent foundation operating under an open-source model. In March, Nvidia introduced its enterprise-oriented NemoClaw platform, with CEO Jensen Huang stating that companies may, in the future, require their own OpenClaw-like solutions. He also described OpenClaw as “an operating system for personal AI,” suggesting that it could help open up the next phase of AI development.
At the same time, over a dozen major Chinese tech companies rolled out compatible products and related services within weeks, with several cities introducing targeted subsidy policies. By contrast, Europe has yet to see a comparable wave of productization led by domestic firms, nor has there been a significant policy response. As a result, a project created by a European developer appears to have been rapidly absorbed into the broader ecosystems of U.S. and Chinese capital and platforms, while Europe’s presence in this wave remains limited.
In fact, Europe is not a latecomer in the field of AI agents. In academic research, as early as the 1990s, British scholars Michael Wooldridge and Nicholas Jennings had already begun systematically exploring agent theory and applications. During the European Conference on Artificial Intelligence held in Amsterdam in 1994, the two scholars helped consolidate what had previously been a fragmented set of topics into a more coherent research direction. Their 1995 survey paper remains among the most frequently cited works in the field.
On the policy side, EU institutions had also begun addressing AI agents prior to the emergence of OpenClaw. In March 2025, the European Commission formally adopted the term “AI agents” in its policy discussions. In July of the same year, guidance documents related to prohibited practices under the AI Act explicitly noted that children may develop dependencies on AI agents and could be more vulnerable to manipulation, calling for enhanced protections for such groups. In October 2025, the European Commission’s Apply AI Strategy further proposed developing technical and policy toolkits to support the adoption of generative and agent-based AI solutions in public administration.
Nevertheless, it is clear that Europe’s overall pace in this wave appears slower than that of both the United States and China. This contrast also highlights deeper structural challenges facing Europe in the governance and industrial development of artificial intelligence.
How, then, has the European Union responded to the global surge in AI agents triggered by OpenClaw? What differences exist among member states? What structural issues do this wave reveal, and how might they shape the future of AI development in Europe? This article addresses these questions from four perspectives: current deployment, regulatory responses, structural constraints, and future outlook.
Before joining OpenAI, Steinberger had long lived in Vienna and London, where the early development and operation of OpenClaw took place.
Source: OpenClaw / Peter Steinberger
Original link:
https://mp.weixin.qq.com/s/F6xJR5RgATUWByuxCIwchw
