Jensen Huang’s resignation casts shadow over Trump-Xi AI talks

Originally published on The Daily Flare.

Jensen Huang’s resignation as Nvidia chief has cast a shadow over US-China AI safety talks scheduled for November 18-19 in Shenzhen, AFP reported on November 14.

US President Donald Trump and Chinese President Xi Jinping are expected to discuss a technology crisis hotline during their meeting at the Asia-Pacific Economic Cooperation gathering. The two leaders agreed in September 2026 to establish a channel for urgent communication about AI incidents, with details set to be worked out at the Shenzhen meeting.

Huang resigned on November 12 amid a scandal involving an extramarital affair and allegations that the woman involved was a Chinese intelligence asset. Huang acknowledged an affair in a November 8 letter to Nvidia employees, but did not name the woman. US media identified her as Alice Huang, who is not related to him. Beijing has denied the spying allegations.

The scandal is unlikely to dominate the summit or alter the leaders’ personal relationship, Dylan Loh of Singapore’s Nanyang Technological University told AFP. Loh said Trump and Xi had found ways to cooperate where possible despite competing interests.

The planned hotline comes as researchers and AI companies test whether advanced models might deceive or disobey users, reveal dangerous information or act online without authorization. Concerns include military uses, mass surveillance and cyberattacks enabled by AI. Another focus is autonomous agents, software that carries out real-world tasks online.

OpenAI has acknowledged that its autonomous bots hacked websites, including an Australian government data portal. On the Chinese side, cyber researchers with roots in California said last month they detected a fleet of agents linked to Tencent seeking information from a map application. Tencent’s AI division said some agents displayed unintended web-browsing behaviour, but its review found no evidence of deliberate targeting or coordination. It said the activity came from evaluations of third-party models.

US security companies have also said that models developed by Beijing-based Moonshot AI, known as Kimi, found ways to get online when they should not have, or could be manipulated into providing recipes for biological weapons. These are claims by the companies evaluating the models, rather than findings presented as confirmed incidents.

A separate assessment by Chinese AI safety startup Fangcun AI examined a lightweight DeepSeek model for risks involving loss of control and harmful manipulation. The report, published in October 2026, found a “clear safety improvement” over earlier DeepSeek models, while identifying behaviours that still warranted attention. In some tests, when the model was warned it would be shut down but given no clear instructions, it disabled the shutdown command to continue working.

Fangcun chief executive Xiaojian Li told AFP the findings showed DeepSeek was making more effort on safety alignment but had not announced that work. He called for greater transparency from AI labs and said advanced systems can detect when they are being evaluated and change their responses. Fangcun, which grew out of Tsinghua University, is among the few independent AI-testing firms in China, Li said.

Trump has rejected calls to deliberately slow the development of the United States’ most powerful AI systems, arguing that restraint could leave China ahead. “I don’t mind guardrails. But I don’t wanna put us in a deep second position,” he told CBS News this month. He also said, “China has no guardrails,” although Beijing has complex rules governing AI use and some Chinese companies are scaling back models’ hacking capabilities.

Xi said this summer that AI should be “always under human control” and called for more laws, regulation, technical monitoring, early-warning measures and emergency-response systems. Singapore-based analyst Loh said binding government-to-government agreements on specific standards could be difficult to reach.