Quote · The a16z Show
Sriram Krishnan on Open Source AI's Biggest Week Yet
Where this was said
Open-Weight Models & Security: Linus's Law Meets the AI Age
At 19:00 · chapter starts 17:30
Krishnan grew up on open source and is a genuine believer in its security advantages. [1] — Sriram Krishnan "Linus Torvalds said 'given enough eyes, all bugs are shallow.' Sriram Krishnan applies this to AI: open-weight models downloaded from Huggi…" 17:30 He invokes Linus's Law — 'given enough eyes, all bugs are shallow' — to argue that open-weight models are more secure than closed ones because anyone can download them from Hugging Face and inspect every layer. You simply cannot do that with a closed API model. But he immediately counterbalances this with a sobering real-time data point: on the day of the recording, Hugging Face reported an active incident where an AI LLM agent was being used to systematically probe and attack their systems. The uncomfortable irony: the answer to AI-powered cyber attacks is almost certainly AI-powered defense — which requires defenders to have access to the best models. If American frontier model safety guardrails make those models less useful for security work, defenders are already at a disadvantage. [2] — Sriram Krishnan "Hugging Face attacked by AI agent: At the time of recording, Hugging Face reported an active incident where an AI LLM agent was being used …" 18:30
Linus Torvalds said 'given enough eyes, all bugs are shallow.' Sriram Krishnan applies this to AI: open-weight models downloaded from Hugging Face can be inspected by the entire world, making them inherently more auditable than closed frontier models. Security through transparency, not obscurity.
Open-weight models downloaded from Hugging Face can be inspected, fine-tuned, and modified by the entire global community, making them inherently more auditable than closed models.
At the time of recording, Hugging Face reported an active incident where an AI LLM agent was being used to probe and attempt to breach their systems.
Every AI model from the original GPT onward was trained by distilling human knowledge from the internet. The controversy around Chinese models distilling from American ones ignores the fact that distillation is foundational to how all models are built — the real issue is the asymmetric legal playing field.
