Nvidia-led Open Secure AI Alliance pivots to rapid security standards

The alliance, managed alongside the Linux Foundation, is prioritizing practical protocols for incident disclosure and blame-free post-mortems. While the initial guidelines focus on foundational safety, the group is aggressively cataloging open-source contributions from members to build a unified defense against AI-specific vulnerabilities. Companies like Nvidia, Amazon, and Red Hat are already contributing tools ranging from Garak vulnerability scanners to Cedar authorization frameworks.

Despite the rapid growth and the presence of industry titans like Microsoft, Cisco, and Intel, a few high-profile players remain absent. Anthropic, OpenAI, and Google have yet to join the formal alliance, even though the latter two signed the initial open letter urging the White House to protect open-source AI development. The group's formation was catalyzed by recent concerns regarding potential U.S. restrictions on Chinese open-weight models. Proponents argue that building an open, robust infrastructure is the most effective strategy to secure the domestic AI landscape against both external and internal threats.