When even US lawyers are comfortable with Chinese AI, it’s a sign that companies’ fears about security and censorship are abating.
Legal AI startup Harvey on Tuesday announced a new in-house model, built in part using Moonshot’s Kimi K3. And data company Snowflake added models from DeepSeek and Z.ai to its platform, allowing customers to have tasks automatically routed to those models when they’re the cheapest or best-performing option.
“We see a lot of value out of it, besides just the cost,” Mayank Upadhyay, Snowflake’s head of security and trust, told journalists in New York on Tuesday. Since Snowflake is hosting the models and has put guardrails on them, there’s less concern over data privacy or that the models’ answers will be influenced by Beijing’s worldview, he added. “I’m not going to these models to ask them cultural questions. I’m doing some sort of data processing and it’s fairly objective.”
Sports merchandiser Fanatics’ VP of data and AI said measuring the ROI of AI tools has become central to her job, and that on Chinese models, “the cost aspect alone demands attention.” Businesses “have to have a stance on it that’s not just a blanket, ‘No because it comes from China.’”
