Nvidia chief executive Jensen Huang is meeting US officials and lawmakers in Washington this week amid reports that its export-controlled processors have been used to train advanced Chinese artificial intelligence models.
Huang met with US Commerce Secretary Howard Lutnick on Tuesday, US outlet Axios reported, citing an anonymous source. Neither the US Commerce Department nor Nvidia confirmed that the meeting occurred, and its purpose was not disclosed.
“Jensen is in DC to meet with leaders on both sides of the aisle to discuss Nvidia’s work to strengthen the country’s supply chain by producing US$500 billion in American technology over four years and American leadership in AI, including leading in open source,” an Nvidia spokesperson said.
Huang’s arrival in the US capital comes after senior Trump administration officials last week accused Beijing-based start-up Moonshot AI of using export-controlled Nvidia chips in Thailand to train its latest models, including the groundbreaking Kimi K3.
US export controls restrict the sale of Nvidia’s most advanced processors to China.
Critics of the company argue that some of the chips have nevertheless reached Chinese users through alleged smuggling networks, overseas data centres and intermediaries.
The US Bureau of Industry and Security (BIS) said last week that it was investigating the allegations involving Moonshot.
On Tuesday, tech outlet The Information reported that Moonshot is seeking access to more of Nvidia’s advanced Blackwell chips to train its next flagship model, Kimi K4, citing anonymous sources.
Moonshot and Nvidia did not respond to queries about the use of Blackwells to train Moonshot’s models.
The release of Kimi K3 sparked alarm in the US, as some analysts claimed it narrowed the gap between the US and China in their heated AI race.
Huang’s visit also comes as Washington debates a separate but related question: whether Chinese open-weight models should face restrictions in the US.
Nvidia has emerged as a prominent opponent of such limits, arguing that open development is important to American leadership in AI.
Kimi K3’s open-weight nature, which allows developers to download and modify the model, contrasts with the largely proprietary approach taken by leading US companies such as OpenAI and Anthropic.
Its release has intensified debate over whether the Trump administration could restrict Chinese open-weight models in the US – a move opposed by many American start-ups.
Nvidia has spearheaded efforts in recent days to rally the US tech industry in support of open-source AI development, including a public letter opposing a ban on open-weight models and the establishment of a new industry coalition, the Open Secure AI Alliance, to promote collaboration on AI safety.
The Trump administration is expected to unveil a new regulatory framework for governing advanced AI systems this week, though it remains unclear whether it will touch upon open-weight models.
During a podcast appearance on Tuesday, OpenAI CEO Sam Altman said his company was committed to open-weight models and claimed that OpenAI’s models are superior to Kimi K3.
He also dismissed concerns that cheaper Chinese open models will disrupt OpenAI’s business model by eating into its margins, arguing that its user base is large enough that even modest margins are sufficient for the company to reach trillions of dollars in revenue.
“This is not in, like, my top 10 list of worries,” he said.
At the centre of growing debate in the US about AI governance and regulation is the technology’s rapidly advancing capabilities, which are worrying policymakers and industry practitioners alike.
On Tuesday, more than 1,000 employees of leading American AI companies, including Google, OpenAI, and Anthropic, issued a public statement calling for Washington to develop the technical and governance tools to “deliberately pace the frontier of automated AI development”. -- SOUTH CHINA MORNING POST
