Open Source AI Is 12 Months Behind the Frontier—Here Are Three Escape Routes

The author argues open-source models lag the frontier by 12 months, predicting a similar security backlash next year. Three escape routes: frontier models desensitize security apparatus, convince decisionmakers benefits outweigh costs, or a research program to make open-source as safe as closed models, with examples.

• 开源模型
• AI安全
• 安全策略
• 前沿模型