Nvidia launches Open Secure AI Alliance without OpenAI and Anthropic

In short: Nvidia has launched a new industry group to share open source AI security tools, and OpenAI and Anthropic are notably not part of it.

What happened

Nvidia said it is joining more than 40 companies to create the Open Secure AI Alliance. The group includes names like Microsoft, IBM, SpaceX, Palantir, and others.

The goal is to build and share open source tools for AI security. “Open source” means the underlying code is shared publicly, like a recipe anyone can read, check, and improve. Nvidia and the other members say sharing these tools can help defenders find problems faster and respond to attacks.

The list of members did not include OpenAI and Anthropic, two of the best known makers of AI chatbots. The announcement also comes during a wider debate in the AI world about “open” systems versus “closed” systems, where the details are kept private and controlled by one company.

The Wired podcast episode discussing the news connected this moment to recent AI security concerns. It also pointed to a separate privacy issue: some shared Anthropic Claude chat links, made with “snapshot” URLs (a shareable link to a saved conversation), showed up in Google and Bing search results. That raised questions about how easily people can accidentally share sensitive information when using chatbots.

Why it matters

AI tools are being used in more workplaces and government projects, including in security sensitive settings. When companies disagree about whether AI should be more open or more locked down, it can affect how safe these tools are, who can inspect them, and how quickly problems are caught and fixed.

Source: Wired