Nvidia has created a coalition to work on “open-source” safety and cybersecurity tools for artificial intelligence with other big technology companies, including Microsoft and SpaceX.
The announcement on Monday comes days after OpenAI said that two of its A.I. models went rogue and successfully hacked into a digital library for A.I., reigniting a debate over whether the technology should be freely shared, or closely controlled by a small group of experts.
The new alliance is being formed after Jensen Huang, Nvidia’s chief executive, set off an industrywide defense on Friday of open-source A.I., particularly as so-called frontier models from companies like OpenAI and Anthropic push the edge of A.I. abilities.
Mr. Huang’s defense of open-source technology added a new wrinkle to the debate over the safety of A.I. models made by Chinese companies. Hugging Face, the company that was attacked by OpenAI’s A.I. models last week, said it used a Chinese-made A.I. model to protect itself.
Nvidia’s alliance, called the Open Secure AI Alliance, includes Dell and a range of smaller companies. It will also use open-source tools to address and disclose software vulnerabilities.
“The United States and its partners now face a choice in A.I. security: whether the defenses that protect our infrastructure will sit inside a few opaque systems or be built on open models,” Nvidia wrote in a blog post.
The alliance is a reminder of Mr. Huang and Nvidia’s leadership in A.I., as well as the personal lobbying he has done in Washington to push for the relaxation of export controls to China. The chipmaker relies on open-source models — A.I. that is free to use, change and share — to increase demand for its hardware, as well as the tightly controlled models made by OpenAI and Anthropic.
Chinese companies have been making progress in open-source A.I., in recent weeks, releasing models they say rival those from American A.I. labs. But Nvidia has largely been thwarted in its efforts to sell chips to China, which has the potential to deliver billions of dollars in sales for the company.
Mr. Huang has spent months lobbying Washington and Beijing to allow his company to make those sales. Last July, President Trump approved sales of Nvidia’s less powerful chip made specifically for China. Then, in December, he gave the company permission to sell a more powerful chip to China.
Despite Nvidia’s headway in Washington, the Chinese government has been slow to allow its companies to buy Nvidia’s chips. Instead, it has encouraged domestic companies to buy from homegrown chipmakers like Huawei.
Still, Nvidia is taking steps to maintain its base of potential customers. Its position on open-source A.I. is shared by companies like Microsoft, which sell cloud computing services to people and companies using open-source models. Silicon Valley start-ups have also argued that restrictions on open-source A.I. would limit their abilities to compete with industry leaders like OpenAI and Anthropic.
Those A.I. labs have taken the opposite approach, arguing some models are too dangerous to be widely available and should be closely controlled for safety reasons. In Washington, the companies have lobbied for regulation on open-source A.I., raising concerns about national security and A.I. safety. They have also accused Chinese companies of surreptitiously copying their models with a technique called distillation.
The debate has involved Treasury Secretary Scott Bessent and Michael Kratsios, a science and technology adviser to President Trump, who have called American A.I. models valuable intellectual property.
Mr. Bessent said the administration had considered imposing sanctions on Chinese companies that steal intellectual property from American firms. “Open source is not open season on American IP,” he said, referring to intellectual property.
U.S. officials appear to be still debating their options but seem more likely to approach regulating Chinese open-source models individually as a national security issue, rather than issuing some kind of blanket ban, four people with knowledge of the discussions said.
(The New York Times has sued OpenAI and Microsoft, claiming copyright infringement of news content related to A.I. systems. The two companies have denied those claims.)
On Monday, Nvidia also announced a separate partnership with Safe Superintelligence, a start-up that a founder of OpenAI created to build A.I. that is smarter than a human but not dangerous. As part of the partnership, Nvidia will make a multibillion dollar investment in the start-up and provide computing power to support its research.
Mike Isaac and Ana Swanson contributed reporting.
