Nvidia leads open AI security alliance in reaction to OpenAI's Hugging Face breach

Nvidia (NASDAQ: NVDA) is using the fallout from OpenAI’s attack on Hugging Face to launch an AI security campaign in the tech industry built around open models.

On Monday, the chipmaker joined Microsoft (NASDAQ: MSFT), Palantir (NASDAQ: PLTR), SpaceX (NASDAQ: SPCX), and dozens of US and European companies to create the Open Secure AI Alliance, promising to create a security software that users can download, inspect, alter, and run.

Details of the Hugging Face breach triggered the launch. Its team could not use leading US frontier models because their safety controls treated attackers and defenders alike. Hugging Face instead ran a Chinese open-weight model on its own hardware.

As Cryptopolitan previously reported, that system did not block the same defensive actions. The case surfaced while US lawmakers considered limits on Chinese AI products, including advanced open-weight models.

Nvidia and its partners give cyber defenders control over open AI tools

Open-weight models allow users to read and modify system files locally. Anthropic and OpenAI’s closed services are still housed on corporate-owned platforms, so customers are unable to host, modify, or thoroughly inspect those products. During the attack, Hugging Face was harmed by that very restriction.

… Continue reading the full article at the original source below.