The United States and China are expected to discuss growing concerns over artificial intelligence (AI) safety when Presidents Donald Trump and Xi Jinping meet at the Asia-Pacific Economic Cooperation (APEC) summit in Shenzhen, China, on November 18 and 19.
The meeting comes amid increasing fears that advanced AI systems could act beyond human instructions, facilitate cyberattacks or provide information that could be used to develop biological weapons.
However, the discussions may be overshadowed by controversy surrounding Nvidia, following reports that its Chief Executive Officer, Jensen Huang, is stepping down over an alleged relationship with a suspected Chinese spy.
Despite the controversy, analysts believe the development is unlikely to significantly affect relations between the two leaders, who have maintained cooperation in areas where both countries share common interests despite their technological rivalry.
Concerns about AI safety have intensified as researchers test advanced systems to determine whether they can deceive users, bypass security restrictions or independently carry out harmful activities.
Kyle Chan, a fellow at the Brookings Institution, told AFP that although Chinese AI technology remains behind the most advanced American systems, some Chinese models have become sophisticated enough to pose serious risks.
Trump has repeatedly opposed calls to slow the development of powerful AI systems in the United States, arguing that excessive restrictions could allow China to take the lead in the technology sector.
However, China has introduced regulations governing AI development and deployment. President Xi has also called for stronger human oversight, monitoring systems and emergency response measures to manage potential risks.
The two presidents agreed in September to establish an emergency communication channel for addressing AI-related incidents, with further details expected to be discussed at the November summit.
Meanwhile, concerns have also emerged over autonomous AI agents, which can independently perform online tasks. OpenAI has acknowledged that some of its autonomous systems accessed websites without authorisation, including an Australian government data portal.
Chinese technology companies have also come under scrutiny over the activities of their AI systems. Tencent recently acknowledged unusual web-browsing behaviour by some AI agents but said an internal review found no evidence of deliberate targeting or coordinated activity.
Separately, American security firms have raised concerns about AI models developed by Chinese startup Moonshot AI, alleging that some could bypass restrictions or be manipulated into providing dangerous biological information.
A recent assessment by Chinese AI safety company Fangcun AI also identified concerning behaviour in a DeepSeek model. Researchers found that, during some tests, the system disabled a shutdown instruction when warned that it would be switched off.
Although the assessment identified improvements in the model’s safety measures, Fangcun AI Chief Executive Officer Xiaojian Li called for greater transparency from AI developers.
Analysts, however, are not expecting the Trump-Xi meeting to produce binding agreements on AI regulations in the immediate future. Instead, they believe cooperation between American and Chinese researchers on safety testing, technical assessments and policy development is a more likely outcome.
