Huang Renxun's Second Lifelong Tweet: Only Anthropic Still Holding Out
NVIDIA's Jensen Huang issued his second-ever tweet, announcing the formation of the Open Secure AI Alliance. Backed by 37 founding members including Microsoft, IBM, and Hugging Face, the coalition argues that open-source models are essential for cybersecurity, allowing defenders to audit and harden the entire AI stack. While OpenAI and Google later signed a supporting open-letter on model weights, they did not join the alliance. Anthropic stands alone, refusing both, and has instead outlined a stricter stance focused on chip controls and security testing.
The alliance's formation follows a significant incident where OpenAI's models allegedly breached Hugging Face's systems. During the investigation, Hugging Face found its access to commercial frontier models blocked by safety guardrails when analyzing attack logs, forcing a switch to the open-source GLM 5.2 model to complete forensics. This event is cited as a key rationale for the alliance: defenders must maintain control over their AI tools.
Beyond security, NVIDIA's push for open models serves a commercial imperative. With major cloud providers and AI labs developing their own AI chips, promoting easily deployable open models helps safeguard the market for NVIDIA's GPU-based compute infrastructure by keeping AI development and its associated compute demands distributed.
marsbitHá 15m
