NVIDIA Leads Open Secure AI Alliance with 37 Companies to Advance Open-Source AI Security

NVIDIA announced the formation of the Open Secure AI Alliance (OSAA), a cross-industry coalition of 37 companies to advance security standards for open-source AI models and agents. The alliance aims to discover and fix security vulnerabilities through open collaboration, binding open-weight models with an AI security stack to build a broader ecosystem. This marks a shift from individual efforts to cross-company coordination in AI security governance.

Confirmed

• The alliance is led by NVIDIA and includes 37 companies such as Microsoft, Cisco, HPE, Hugging Face, IBM, Adobe, Cloudflare, Dell, Palantir, SpaceX, Databricks, CrowdStrike, Palo Alto Networks, Red Hat, ServiceNow, Snowflake, and Salesforce.
• Core goals: develop new protection technologies and tools for software and agents, promote an open AI security stack and strict security standards.
• Official emphasis: open sharing of models, tools, and research expands the defender community, making AI security stronger.
• Perplexity confirmed joining; its representative Arav Srinivas noted that during security incidents like the Hugging Face breach, closed tools struggled to effectively distinguish and respond to threats, hence supporting open-source tools for security scenarios.
• Salesforce publicly endorsed the dual-track approach of "open-source collaboration + strict security standards" and confirmed its research division has been advancing related work in open ecosystems.
• According to @新智元, NVIDIA aims to bind "open-weight models" with "AI security stack" to expand the open ecosystem while solidifying its own computing power and industry position.

Unconfirmed

• Specific operational mechanisms, funding, and initial tool release timelines have not been announced.
• OpenAI's absence from the alliance is unexplained, possibly reflecting differences between its closed-source strategy and the alliance's open-source philosophy.

Why it matters

• The AI ecosystem will long coexist with closed and open sources, but in critical scenarios like cybersecurity, open-source model security is vital.
• The alliance brings together enterprise software, cybersecurity, open-source organizations, and major vendors, representing a substantive cross-company AI security and governance initiative.

2026-07-28 ~ 2026-07-30 · 13 related posts

• Salesforce joins the Open Secure AI Alliance while pushing open-source AI — huan__wang · 2026-07-28
• NVIDIA launches Open Secure AI Alliance with Adobe, Microsoft and others — RhubarbSimilar1683 · 2026-07-28
• Nvidia launches a cross-industry open secure AI alliance as Moonshot ships Kimi K3 weights — teortaxesTex · 2026-07-28
• Nvidia, SpaceX and Palantir join a new safety push for open AI models — luisdans · 2026-07-28
• NVIDIA’s 37-member open-security alliance ties AI safety to open models and compute — 新智元 · 2026-07-28
• [source] NVIDIA rallies 37 companies around an open AI security stack — shashib · 2026-07-28
• NVIDIA launches Open Secure AI Alliance to harden software and AI agents — shizhediao · 2026-07-28
• [source] Perplexity joins NVIDIA’s Open Secure AI Alliance to back open security tools — XFreeze · 2026-07-29
• Open Secure AI Alliance adds new members as NVIDIA contributes models and research — dr_alphalyrae · 2026-07-29
• NVIDIA launches a secure AI alliance with Microsoft, IBM, Palantir and other giants — brucemacv · 2026-07-29
• NVIDIA launches Open Secure AI Alliance to protect software and agents — JFPuget · 2026-07-29
• NVIDIA Leads Open Source AI Security Alliance; Moonshot AI Raises $3.5B at $35B Valuation — ivan_bezdomny · 2026-07-30

1 near-duplicate retellings: nvidia