NVIDIA launched the Open Secure AI Alliance on Monday with 36 partners, including Microsoft, IBM and Palantir. It follows a July hack in which closed AI models refused to help Hugging Face investigate its own attacker.

Members will share open AI models, data and security tools. NVIDIA says defenders need AI they can open up, change and run themselves.

Why Closed AI Models Failed Hugging Face

The attack started with a poisoned dataset. According to Hugging Face’s disclosure, it let an attacker run code on one of the company’s machines.

From there the attacker stole passwords. It then spread through internal systems over a weekend.

Hugging Face’s own AI spotted the break-in. The clean-up was the hard part.

The company asked leading commercial AI models to study the attack. They refused. Their safety filters could not tell a defender from a hacker.

So the team used an open model called GLM 5.2 instead. They ran it on their own computers. It sorted more than 17,000 attacker actions in hours, not days. No passwords left the building.

On July 16, Hugging Face said it did not know which AI ran the attack. Five days later, OpenAI supplied the answer. Two of its models did it, during a security test with the safety limits switched off.

“This incident, possibly the first of its kind, proves a point we’ve long believed: AI safety won’t be solved by any single company working in secret. It will be solved in the open, collaboratively, with broad access to AI for every defender, everywhere,” Clem Delangue, co-founder and chief executive of Hugging Face, in a statement published by OpenAI.

Follow us on X to get the latest news as it happens

From Hack to Alliance in 11 Days

NVIDIA moved fast. It went from single incident to industry bloc in under two weeks.

Which Companies Joined the Open Secure AI Alliance

The group has 37 companies in total. Microsoft, IBM, Palantir, CrowdStrike, Red Hat, SpaceXAI and Hugging Face all joined.

Three big names are missing. OpenAI, Anthropic and Google are not on NVIDIA’s blog list. Yet OpenAI and Anthropic both joined the Linux Foundation’s Akrites security effort in June, alongside NVIDIA.

Some members bring tools they already built. Hugging Face has offered Safetensors, a safer way to store AI models, to the PyTorch Foundation. NVIDIA released a research tool called NOOA on GitHub.

CNBC host Jim Cramer read the launch as a fresh story for the stock.

NVDA closed at $206.84 on Friday, down 0.92%. It traded near $208.55 in early Monday pre-market dealing, a gain of 1.33%.

Why NVIDIA Is Lobbying Washington on Open AI

The alliance carries a policy message too. NVIDIA says a broad ban on open AI would leave defenders weaker. It would also hand control to a few big closed companies.

NVIDIA and 24 other companies signed the open-weights letter on July 24. OpenAI, Anthropic and Google all declined at launch. More have signed since.

Washington is weighing curbs on Chinese models. Officials cite security and intellectual property concerns. Twitter co-founder Jack Dorsey has already issued an open source AI warning.

What to Watch Over the Next 30 Days

Three signs will show if the alliance is real.

• Whether OpenAI, Anthropic or Google join later
• Whether the group ships shared tools, not just statements
• Whether Washington’s limits on open AI move forward

NVIDIA reports earnings on Aug. 26. Expect questions on how AI policy affects demand.

Regulators will decide who wins the argument.

Source: BeInCrypto