Arcee, a U.S. open-model lab, is pushing back on claims that Chinese open-weight AI models are inherently dangerous for American companies to run. TechCrunch reports that Arcee CTO Lucas Atkins argues these models should be treated as software that needs inspection and testing, not as tools that automatically give their creators access.
The debate has intensified as Chinese models such as Moonshot AI’s Kimi K3 and Alibaba’s Qwen become more capable and cheaper to run than many closed U.S. alternatives. Some policymakers and proprietary AI companies have raised concerns about security and competition, including the possibility of bans.
Atkins says the remote-control analogy is wrong. If a company downloads and runs model weights in its own environment, the model maker does not automatically have access to that deployment. He still says enterprises should put any model through security review, post-training, bias and toxicity checks, and testing for sensitive use cases before using it.
The argument is notable because Arcee could benefit from fear around Chinese models. Its position is narrower: open-weight models are not risk-free, but the risk should be evaluated technically rather than assumed from country of origin alone.