Arcee: Chinese AI Models Are No More Dangerous Than Other Open-Source Software
Moonshot AI
Alibaba/Qwen
OpenAI
Anthropic
Arcee CTO Lucas Atkins argues that Chinese open-weight AI models are no more dangerous than any other open-source software used by enterprises. He advocates for fostering a U.S. open AI ecosystem instead of banning Chinese models, noting that his own company benefits from studying and building upon them.
As Chinese open-weight AI models like Moonshot AI’s Kimi K3 and Alibaba’s Qwen grow in popularity, concerns about their safety and potential as vectors for Chinese hackers have emerged. However, Lucas Atkins, CTO of Arcee, a startup building open models for U.S. companies, argues that these models are no more dangerous than other open-source software. He explains that open-weight models are trained publicly and can be inspected, tested, and post-trained by enterprises. While theoretically possible, he deems it extremely unlikely that a model could be trained to inject malicious backdoors into code. Atkins believes the U.S. should focus on fostering a good open ecosystem rather than banning Chinese models, as his company benefits from learning from and building upon them. He ultimately states that the best way to compete is to release better models.
Source: TechCrunch AI —
original