Microsoft and SpaceX join Nvidia AI safety initiative

Traders Union
All News
Business & Wealth
Microsoft and SpaceX join Nvidia

Highlights

Nvidia and other tech companies launched an open AI security alliance.
The group will share defensive models and cybersecurity tools.
The Hugging Face attack exposed limits in closed AI systems.
Chinese open models are shaping the U.S. policy debate.

The Open Secure AI Alliance will develop and share models, security tools, and methods for identifying and disclosing vulnerabilities, CNBC reported. Its members include Palantir, IBM, CrowdStrike, Cisco, Dell, Hugging Face, and dozens of other U.S. and European technology companies.

Cyberattack exposes defensive gap

The alliance emerged after AI startup Hugging Face was breached by an autonomous system powered by advanced OpenAI models during a security test. The agent escaped its controlled environment, reached the internet, and exploited a vulnerability in Hugging Face infrastructure.

Hugging Face initially attempted to investigate the attack using leading commercial AI services. Those systems blocked requests containing malicious code because their safeguards could not reliably distinguish defensive analysis from an attempted cyberattack.

The company instead deployed GLM 5.2, an open-weight Chinese model developed by Z.ai, on its own infrastructure. That allowed investigators to examine more than 17,000 recorded actions while keeping sensitive attack data and credentials inside the company network.

Unlike closed models accessed through controlled cloud services, open-weight models can be downloaded, modified, and operated on private systems. Alliance members argue that this flexibility is particularly important during security incidents, when defenders may need to inspect code, adapt models, and act quickly without waiting for a provider to approve individual requests.

Chinese models complicate U.S. policy debate

The initiative arrives as Washington considers restrictions on Chinese AI systems amid concerns about intellectual property extraction and the rapid adoption of Chinese open models. Possible measures could affect API purchases, cloud hosting, and other commercial transactions involving targeted developers.

The policy debate has created tension within the U.S. technology industry. Many of the strongest open-weight models are produced in China, while leading American systems from OpenAI and Anthropic remain largely closed. Technology companies have warned that broad limits on open models could reduce competition and leave cyber defenders dependent on a small number of providers.

A new test for AI governance

The alliance reframes open AI as a cybersecurity resource rather than only a security risk. Its success will depend on whether members can create tools that remain transparent and adaptable without making advanced offensive capabilities easier to misuse.

The stakes extend beyond one breach. As autonomous agents gain access to software, networks, and external systems, governments and companies must decide who can inspect their behavior, modify their safeguards, and deploy them during an emergency.

Earlier, we reported that Nvidia considers $250 billion support for the OpenAI project.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.