NVIDIA just announced Open Secure AI Alliance with goal to build and share open tools that promote responsible use of and trust in AI
Read more: https://blogs.nvidia.com/blog/open-secure-ai-alliance/?ncid=so-twit-957725
Main ponts ->
- NVIDIA launched the Open Secure AI Alliance (July 27, 2026) ~35 partners including Microsoft, IBM, Red Hat, Cisco, CrowdStrike, Hugging Face, Palantir, the Linux Foundation, SpaceXAI
- Goal: build shared open models, harnesses and tools for cyber defense instead of leaving AI security inside a handful of closed systems
- Trigger case: in the recent Hugging Face breach, closed AI tools blocked forensic work because they couldn't tell attacker from defender;; HF used the open-weight GLM 5.2 on its own infra to analyze 17k+ actions and contain it
- Counter to "open = unsafe": misuse risk is real but doesn't vanish with closed weights; answer is safeguards + evaluation + fast remediation, not locking defenders out
- Contributions: NVIDIA's NOOA agent framework (GitHub), Hugging Face donating Safetensors to PyTorch Foundation, Microsoft MDASH, SpaceXAI open-sourcing Grok Build (Grok weights planned) etc
- Policy ask: treat open models as defensive assets, not liabilities, blanket restrictions would concentrate dependence in a few closed providers