While using Chinese technology might seem counterintuitive to Europe’s goals, running open-weight Chinese models on local servers can actually give firms greater operational control than relying on proprietary US services, according to a Germany-based consultant.
“A Chinese-developed open-weight model operated on European infrastructure, with data remaining under the company’s control, may in some respects offer greater operational sovereignty than consuming a proprietary foreign [model] that can be changed, repriced or withdrawn remotely,” said Volker Pfirsching, a Munich-based partner at management consultancy Arthur D. Little, in an interview this week.
“Sovereignty should not simply be equated with the nationality of the supplier,” Pfirsching said.
Because many Chinese models are open-weight – allowing developers to download, modify, and host the underlying code locally – they give enterprises flexibility over where and how the technology is deployed.
That control lies at the heart of Europe’s AI debate. Hosting open-weight systems locally keeps sensitive data on European soil and under corporate oversight. However, adopting Chinese foundation models risks introducing fresh supply-chain dependencies, complicating Brussels’ broader push for technological self-reliance.
Yet, for many European firms, the immediate appeal of Chinese models is practical rather than ideological: they offer competitive performance at far lower costs for specific business tasks.
“Companies increasingly realise that using the most powerful frontier model for every task is economically inefficient,” Pfirsching said, noting that routing specific workloads to suitable models was far more cost-effective.
Chinese AI players are already capitalising on this demand.
Last month, Shenzhen-based and Hong Kong-listed Xunce Technology partnered with European digital services provider Lutech to bring its TokenOS platform to Europe and develop an “AI token factory” designed to meet European data regulations.
Major European conglomerates are also experimenting with Chinese foundation models. Industrial giant Siemens was using systems from DeepSeek and Alibaba Group Holding’s Qwen for select applications, according to a Financial Times report last month. Alibaba owns the South China Morning Post.
Despite the momentum, Chinese models face regulatory hurdles in Europe.
Under the European Union’s General Data Protection Regulation, companies transferring or making personal data accessible to China-based entities must implement strict legal safeguards and prove equivalent protection levels, Pfirsching said.
Recent regulatory enforcement surrounding TikTok’s user data transfers to China has further heightened scrutiny.
While self-hosting open-weight models locally helps keep sensitive corporate data within Europe, it does not eliminate the need for cybersecurity assessments, software supply-chain controls, or compliance with the EU AI Act, which came into effect earlier this month, according to Pfirsching.
As a result, European enterprises are gravitating towards a balanced strategy that combines American, European and Chinese AI models to avoid single-supplier risks.
“Chinese models are likely to become a significant part of that portfolio because their performance and efficiency are increasingly competitive,” Pfirsching said.
Looking forward, Europe’s objective is not technological autarky, but ensuring sufficient market alternatives so that no single ecosystem becomes indispensable, he added.
Source: www.scmp.com
No comments:
Post a Comment