Nvidia (NASDAQ: NVDA) is turning the fallout from OpenAI’s attack on Hugging Face into a broader AI security push built around open models.

On Monday, the chipmaker joined Microsoft (NASDAQ: MSFT), Palantir (NASDAQ: PLTR), SpaceX (NASDAQ: SPCX), and dozens of U.S. and European companies to form the Open Secure AI Alliance, which aims to develop security software that users can download, inspect, modify, and run.

The launch came after details of the Hugging Face breach surfaced. According to the report, Hugging Face’s team could not use leading U.S. frontier models because their safety controls treated attackers and defenders alike. Instead, Hugging Face ran a Chinese open-weight model on its own hardware.

As Cryptopolitan previously reported, that system did not block the same defensive actions. The incident emerged as U.S. lawmakers were considering limits on Chinese AI products, including advanced open-weight models.

Nvidia and partners say open AI tools give cyber defenders more control

Open-weight models allow users to read and modify system files locally. By contrast, Anthropic and OpenAI’s closed services remain on corporate-owned platforms, meaning customers cannot host, modify, or thoroughly inspect those products. In the Hugging Face incident, that limitation was part of the problem, and it also helps explain why security and governance debates around open and closed AI systems have become so prominent in Washington and Silicon Valley.

“The Open Secure AI Alliance will work to remediate and disclose vulnerabilities using open technologies,” Nvidia said. “The recent Hugging Face security incident delivered a clear reminder: cyber defenders need open, frontier agentic systems for self-defense.”

Alliance members plan to share fixes created with open technology and publish security vulnerabilities as well. They also want defenders to run AI agents inside private networks and tailor the software to specific threats without waiting for a third-party provider.

Nvidia described the OpenAI and Hugging Face incident as a “practical truth.” Last week, Nvidia, Microsoft, Meta Platforms (NASDAQ: META), Palantir, and more than 20 other companies sent a letter to lawmakers asking authorities not to impose “premature restrictions” on open-weight models that would “stifle competition or push innovation overseas.”

The company also pointed to the fact that Chinese developers currently provide virtually all of the powerful downloadable models, which it said increases the urgency of stricter U.S. regulation.

Nvidia is pairing the security initiative with major infrastructure commitments. It has already signed letters of intent for projects worth more than $500 billion with South Korea’s SK Group. The agreement covers future memory supply, AI factories, and joint development of new memory technology.

Nvidia backs major computing projects in South Korea and Ohio

In South Korea, SK Telecom (KRX: 017670) plans to build a 2-gigawatt AI cloud using SK Hynix (KRX: 000660) HBM4 memory, the Vera Rubin computing system, and Nvidia’s DSX platform. The first facility is expected to open in the first half of 2027.

Nvidia’s DSX full-stack design, which combines accelerated processors, integrated systems, software, networking, and partner equipment, will also be used in the plants. As AI demand grows across Asia-Pacific, particularly in South Korea, companies are seeking ways to reduce energy use and processing costs.

Nvidia is also preparing to provide a cash guarantee of about $250 billion for an OpenAI data center facility in southern Ohio. The Wall Street Journal reported that the project’s total value could approach $500 billion once all chips are included. The site would provide 10 gigawatts of capacity, making it the largest announced data center project.

A SoftBank energy subsidiary is building the campus. Nvidia’s backing could help SoftBank Group (TYO: 9984) borrow on more favorable terms. OpenAI is private, unprofitable, and lacks an investment-grade credit rating, and it has spent several weeks in advanced negotiations to lease the property.

OpenAI has shown the strongest interest in the site. Anthropic, Microsoft, and Alphabet (NASDAQ: GOOGL) have also discussed it with Howard Lutnick. If a lease is signed, OpenAI would become a direct data center tenant for the first time. It currently rents most of its computing capacity from Microsoft, Amazon (NASDAQ: AMZN), and Oracle (NYSE: ORCL).

OpenAI recently raised its planned computing spending through 2030 to about $750 billion, up from roughly $600 billion earlier this year.

Nvidia, now valued at about $5 trillion, is seeking larger long-term chip orders. Its latest annual report warned that data center financing could reduce near-term cash flow and increase exposure to customer credit risk.