OpenClaw founder Peter Steinberger said in a Bloomberg interview that China and the United States are developing sharply different approaches to AI agents. In his view, China has turned into a large-scale testing ground for OpenClaw, with students, office workers, and older users all trying the tool, while some companies have gone as far as requiring employees to use it and report what they automated that day. The picture in the US is very different. OpenClaw has gained strong traction with developers and early adopters, but it has not reached the same level of mainstream enthusiasm, and some companies have already restricted employee use over fears that AI agents could act unpredictably.
China’s rollout is faster, with usage nearing double that of the US
According to Bloomberg, OpenClaw usage in China is nearly twice that of the United States. Steinberger pointed to several expansion efforts: Baidu plans to integrate OpenClaw into its mobile search service, which has about 700 million users; Alibaba has launched “JVS Claw”; ByteDance introduced “ArkClaw” through Volcano Engine; and Tencent Cloud is offering free installation services in 17 cities. The pace is rapid, and the push is coming from multiple large companies at once.
That momentum has not erased official concerns. The report said Chinese authorities have started limiting the use of OpenClaw in state-owned enterprises and government agencies, citing worries over data flows and the security boundaries of open-source tools. The contradiction is hard to miss: the same market pushing AI agents into everyday workflows is also tightening controls in more sensitive settings.
US companies remain cautious as safety concerns stay unresolved
In the US, the resistance is more visible. Steinberger said some companies have moved to restrict OpenClaw because of security concerns. He also raised a recent episode involving a Meta security researcher who was mocked after publicly warning about the risk of agent errors and unintended actions.
Steinberger said, “If people laugh, others will stop speaking up.” His concern is that debate around AI agent safety could shrink if researchers and employees feel they will be dismissed for raising risks. The split between China and the US is not only about speed of adoption. It also reflects very different corporate instincts on efficiency, control, and acceptable risk.
Steinberger joins OpenAI, while OpenClaw Foundation takes shape
Steinberger officially joined OpenAI on February 15, 2026, where he is leading development of Codex agent technology. OpenAI chief executive Sam Altman publicly called him a “genius” and said he would help drive the next generation of personal AI agents. Steinberger said Codex now has more than 2 million weekly users, and described his goal as making the line between programmers and non-programmers less rigid by letting AI agents write and execute code on behalf of users.
Support from major tech companies is also growing. Nvidia chief executive Jensen Huang said at the Morgan Stanley TMT conference in early March that OpenClaw was “the most important software release ever,” adding that every company needs an OpenClaw strategy just as they once needed HTML and Linux strategies. At the same time, Steinberger is working to establish the OpenClaw Foundation, with legal procedures expected to take several more weeks. Nvidia has already joined, ByteDance is also on board, Tencent is expected to follow, and Steinberger said discussions with Microsoft have begun. He said he wants the foundation to serve as a “Switzerland” for open-source AI agents, staying neutral and open rather than becoming tied to any single side.