Huawei's Xu says Chinese AI not powerful enough yet to see frontier risks

Huawei's CEO Eric Xu has expressed skepticism that Chinese AI developers have reached the level of sophistication necessary to encounter the safety risks highlighted by U.S. AI firms. During a speech at the company's annual Connect conference in Shanghai, Xu argued that because Chinese developers lack the same computing power as their U.S. counterparts, they may not yet be able to fully comprehend the potential dangers AI models pose.

Xu's comments come amid growing alarm in the United States over the increasing autonomy and control of AI systems. Major AI developers like OpenAI and Anthropic have begun disclosing unintended or unauthorized model behaviors, while executives have called for a slower pace of development to allow safeguards to keep up. In contrast, China has taken a more proactive approach, treating advanced AI as a powerful yet manageable technology while developing mandatory standards and assessments to mitigate risks associated with autonomous agents.

Xu suggested that Chinese AI developers may need to accelerate their pace to reach a point where they can feel the risks associated with AI model development. He emphasized the importance of striking a balance between driving innovation and managing potential risks, stating that Huawei itself expects autonomous AI agents to dominate global AI traffic by 2035.

The company projects that by that year, agents - AI systems capable of planning tasks, making decisions, and utilizing software or tools with limited human supervision - could account for over 90% of global AI processing traffic, with as many as 900 billion active agents. Huawei has identified agent security and privacy as crucial technologies to address these anticipated challenges, given its dominant position in China's $50 billion AI chip market due to U.S. export controls since 2023.