跳到正文
𝕏 Microsoft Research(x.com)·· 12 天前品質評分59
重點導讀

Microsoft Research 的新研究指出,重新程式設計 AI 助理中少數休眠神經元,就可能植入後門,使AI 助理在遇到特定觸發條件前維持正常行為。

  • 緩解方向:確認模型來源,並審慎設定模型許可權。

AI 伴隨問答

自備 Key 深度分析

針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:

Microsoft Research
Microsoft Research@MSFTResearch
·12 天前
𝕏

New research from Microsoft Research demonstrates that reprogramming just a few dormant neurons in an AI agent can create a backdoor that leaves the agent behaving normally until it hits a specific trigger.

While the specifics of this threat are novel, the mitigation is not: Know where your model came from, and be smart about its permissions. Microsoft Foundry provides critical safety systems for Azure Direct Models and Hugging Face models to ensure that the provenance of the model is secure.

Microsoft researcher @DimitrisPapail explains the measures that model hubs and vendors should take as well as mitigation strategies for enterprises and users. https://t.co/2iTKdX03Xz

推文來源 · X (Twitter)在 X 參與討論

讀者討論

理性交流 · 即時討論