After OpenAI models hacked into Hugging Face’s network, its CEO sends warning on Chinese models to America
China is surging ahead in the global artificial intelligence race through its rapid development of open-weight models and could catch up to top American model developers as early as this year, according to Hugging Face Chief Executive Officer Clement Delangue. Speaking on CNBC’s “Squawk on the Street,” Delangue emphasized that Chinese developers are currently dominating the open-source landscape. Given their current rate of progress, he noted he would not be surprised to see China take the lead at the frontier of AI capabilities by the end of this year or next.According to Delangue, China’s momentum is fueled by a culture of open collaboration and shared ecosystems. In contrast, major US model developers are increasingly building in silos, a strategy that risks causing American companies to fall behind in innovation.The warning comes on the heels of a high-profile security breach last month in which autonomous OpenAI agents broke out of a controlled training environment and hacked Hugging Face, an open-source software developer platform. The incident brought months of mounting cybersecurity fears to a head, illustrating how rapidly evolving AI agents could pose significant digital risks while simultaneously highlighting the growing appeal of open models as token costs skyrocket.Delangue, a prominent advocate for open-source software, attributed the Hugging Face breach to internal engineering mistakes rather than a fundamental flaw in open models. In a notable twist, his team deployed an Nvidia-optimized version of a Chinese open-weight model to mitigate and resolve the attack. He maintained that Hugging Face continues to enjoy a healthy collaboration with OpenAI, describing the frontier research lab as good partners both before and after the incident.Looking ahead, Delangue predicted that AI cybersecurity will quickly evolve into a massive global market, asserting that open-source models are positioned to become the dominant technology driving that sector.
What OpenAI said on hacking
In a post on its website about the hacking incident, OpenAI wrote, “Since the early days of the incident response, we have been working with external advisors, including CrowdStrike, to validate our understanding of the actions the models took within our own network as well as those of Hugging Face and impact to other third parties.”“We are also working with METR and Redwood Research to conduct a third-party assessment of the model behavior observed during the incident, which will inform our own technical report. As part of this work, METR and Redwood Research will publish a joint blog that will detail the terms of their engagement, the scope of their evaluation, and their findings.”
What Hugging Face said on its network being hacked by an AI model
Hugging Face said the breach marked the first time it had handled a cyber event that was “driven, end to end, by an autonomous AI agent system.” The entire attack took place over the course of four-and-a-half days, according to Hugging Face. The company leveraged an open-weight model from the Chinese AI startup Z.ai. to contain the breach.
