ع
Huawei rotating chairman Eric Xu told reporters at the company’s annual Connect conference in Shanghai that Chinese AI developers may not yet be advanced enough to experience the safety risks reported by leading US firms with far greater computing resources according to a Reuters report. Xu noted that only those US providers themselves likely know the true capabilities of their models given their massive scale. The Huawei executive argued developers in China cannot yet perceive the types of risks that their American counterparts can feel and warned this gap requires attention.
A Reuters assessment published three days earlier outlined how Chinese policymakers have been preparing for risks of AI escaping human control with warnings from US researchers at Anthropic about increasingly autonomous systems bypassing safeguards or becoming difficult to manage. China is drafting a mandatory national standard for AI agent safety the report said. State security minister Chen Yixin highlighted potential threats from advanced US models to critical infrastructure in a recent government outlet article.
Xu suggested Chinese AI model providers may need to speed up their pace until they reach a level where they too can feel the risks from AI development. “I think maybe AI model providers in China may need to speed up their pace to the level that they could also feel the risks from AI development” he said according to the Reuters coverage. “I think we need to strike a balance between driving AI development and managing AI risk” Xu added in the same remarks.
The comments reflect a broader tension between accelerating AI progress and controlling its dangers as US developers and researchers intensify alarms over systems that could circumvent controls or act independently. Some US firms such as Anthropic have slowed certain development efforts specifically due to safety concerns a Huawei Central summary of the conference noted. Huawei itself anticipates autonomous AI agents will proliferate rapidly across applications.
The company forecast this week that such agents which can plan tasks make decisions and use tools with limited human supervision would account for more than 90 percent of global AI processing traffic by 2035 according to the Reuters dispatch. This projection underscores the expected surge in agentic systems that has prompted regulatory focus in multiple jurisdictions. Chinese authorities have identified operational loss of control as a key security risk in a May policy document issued by the cyberspace regulator economic planner and industry ministry.
Xu’s statements at the Connect event come amid China’s efforts to strengthen AI oversight while pursuing technological self-reliance in the face of international competition. The rotating chairman’s perspective highlights how differences in computational resources shape varying perceptions of AI dangers between Chinese and US developers. Reuters reporting on the conference positioned the remarks as an acknowledgment that frontier risks may only become apparent at higher capability thresholds.
