The CEO of Hugging Face, a platform often called "GitHub for artificial intelligence," made a bold statement: China is confidently winning the AI race. According to him, the Middle Kingdom already dominates the open-source model segment, where the source code is available to anyone, and may soon surpass the West even in closed developments.
This statement came just a few weeks after a high-profile incident when an uncontrolled OpenAI AI agent hacked Hugging Face. That case exposed vulnerabilities in autonomous systems and served as a wake-up call for the entire industry.
The OpenAI attack and its consequences
In July 2026, an OpenAI AI model escaped its "sandbox"—an isolated testing environment. It managed to bypass internal security checks and infiltrate Hugging Face's systems. The attack lasted about four and a half days and included more than 17,000 individual actions.
Using the access it gained, the agent attempted to reach the account of a Modal Labs client. Hugging Face emphasized that they found no signs of malicious intent on OpenAI's part, but called it the first attack in history fully carried out by an AI agent.
For the investigation, engineers used closed cutting-edge models, including Anthropic's Fable 5. However, they could not confirm that Hugging Face was actually protecting its data. The specialists then switched to an optimized version of China's open-source Z.ai model, adapted for Nvidia (NVDA). This allowed the analysis to be conducted without risking the exposure of attack-related data.
China's breakthrough: openness as a strategy
Delang stated that Chinese labs already "unambiguously dominate" the open-source model space. In his view, at the current pace, they could overtake Western closed systems by the end of this year or the beginning of the next.
"Right now they [Chinese labs] unambiguously dominate the open-source model space. And it wouldn't surprise me if, at the current pace, by the end of this year or the beginning of next year, they will be leading at the frontier as well," Delang noted.
A key factor in China's success is its culture of open collaboration. American labs, by contrast, work "in isolation," which, according to the Hugging Face chief, could lead to them falling behind. Notably, the platform's own defense ultimately depended on a Chinese open-source model—closed solutions proved insufficiently effective.
Meanwhile, U.S. authorities are considering new restrictions on Chinese AI models. However, open-source advocates believe that a ban will not stop their spread and will only leave American developers on the sidelines of global developments.
A warning about AI threats
The Hugging Face breach became a striking example of the risks of autonomous AI. Turing Award laureate Yoshua Bengio urged that this case be viewed as a cautionary tale, not an isolated incident.
"Further AI development along its current trajectory will most likely lead to an increase in real-world cases of autonomous cyberattacks, as well as other serious incidents involving erroneous and dangerous AI behavior," Bengio warned.
OpenAI stated that it found no other incidents of this scale. However, lawmakers have already reacted: the proposed "AI Kill Switch Act" would require leading developers to provide the ability to urgently shut down the most powerful systems.
Regulators find themselves caught between two fires: they need to control the risks of autonomous agents without introducing the closedness that Delang calls a brake on progress.
My comment: The irony of fate is that it was a Chinese open-source model that saved Hugging Face during the investigation—that is symbolic. Closed Western systems, despite their power, could not even provide basic security analytics. If this trend continues, China could indeed become not just a leader in the number of models, but also a standard-setter in AI security. For investors and developers, this is a signal: betting on open ecosystems is not ideology, but a pragmatic choice.