US-China AI safety talks continue despite Nvidia CEO scandal

The fallout from the Huang scandal

The resignation of Nvidia CEO Jensen Huang has sent shockwaves through the tech industry and the halls of government. The departure follows reports of extramarital affairs with staff and a more serious allegation involving a Chinese spy. These events have created a volatile atmosphere for international relations.

On the American side, the scandal is expected to fuel a bipartisan push for stricter regulations. Lawmakers and administration officials are likely to increase oversight on how Chinese entities access high-end chips and manufacturing equipment. This could also reignite the debate over whether the US should implement a total ban on Chinese AI models.

While some observers believe the scandal will destroy any chance of cooperation, others see it as a catalyst for domestic policy. The situation provides a clear justification for the US to tighten the screws on Chinese tech companies while maintaining separate tracks for safety discussions.

Tightening the export net

The US government is exploring several legislative paths to limit China's access to advanced computing power. The Senate version of the National Defense Authorization Act includes the AI Overwatch Act, which would grant Congress direct power over advanced chip exports. Another proposal, the Chip Security Act, would require businesses to track their hardware to prevent smuggling.

Additionally, the Multilateral Alignment of Technology Controls on Hardware (MATCH) Act seeks to coordinate with allies to restrict chipmaking equipment. There is also a focus on closing the "remote access" loophole, where Chinese companies rent chips located in offshore data centers. The Remote Access Security Act would give the Commerce Department more authority to shut down these servers.

The administration has already begun taking action, with reports that President Trump authorized a crackdown on companies diverting Nvidia chips to China. Even with some internal reservations, the Commerce Department is finalizing rules to block Chinese access to restricted chips in countries like Malaysia and Singapore. These moves aim to prevent China from bypassing existing trade barriers.

China's shift in AI priorities

China is also evolving its approach to AI governance, moving beyond simple content moderation. The nation's de facto AI regulator, the Cyberspace Administration of China, has released a report on frontier AI safety. This report notably uses Western-style vocabulary to assess risks like cyber threats and the development of biological weapons.

A key goal for Chinese authorities is ensuring that AI systems remain under full human control. They want to ensure that these models reliably do only what people want them to do. This shift suggests a growing recognition of the technical dangers of advanced AI that transcend national borders.

Despite the tension, China is currently encouraging local suppliers like Huawei to build their own AI stacks. They understand that there will be a capability gap compared to Nvidia chips for a time. However, they view this as a necessary step toward self-sufficiency to avoid permanent dependence on American technology.

The path toward safety dialogue

Despite the political noise, a formal dialogue on AI safety between the US and China remains on track. In September, US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng agreed to establish a formal communication channel. This includes a dedicated incident line for reporting and discussing AI safety emergencies.

Technical discussions are scheduled to take place before the meeting between Presidents Xi and Trump in Shenzhen next week. Secretary Bessent has stated that he wants the two countries to exchange best practices for evaluating and testing models before they are released. He also hopes to establish a framework for keeping human judgment in command of the technology.

The logic here is that AI safety involves common threats that neither country can solve alone. A rogue AI system or a highly capable model in the hands of a dangerous non-state actor poses a risk to the critical infrastructure of both nations. Because of these shared risks, the technical dialogue can proceed even as trade and diplomatic relations face significant strain.