English(EN) A Deterministic Security Runtime for LLM Agents on MPI and Batch-Scheduled HPC Systems: Consensus-Governed, Hardware-Rooted Supervision of Every Rank

新的安全运行时Enclawed增强了HPC系统上LLM代理的安全性

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-27 04:00

研究人员开发了Enclawed，这是一种新颖的确定性安全运行时，旨在提高在高性能计算（HPC）系统上运行的大型语言模型（LLM）代理的安全性。该系统解决了现有HPC环境中受损控制主机的关键漏洞，这可能导致重大的资源浪费甚至危险的后果。Enclawed利用硬件信任根Enclaweder，确保只执行符合策略的操作，并防止恶意行为者伪造关键安全数据。 AI

影响增强了在HPC基础设施上执行关键计算的LLM代理的安全性和可靠性。

排序理由学术论文，详细介绍了HPC系统上LLM代理的新安全运行时。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 9 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，详细介绍了HPC系统上LLM代理的新安全运行时。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Alfredo Metere · 2026-10-27 04:00

面向MPI和批处理调度HPC系统的LLM代理确定性安全运行时：共识治理、硬件信任的每个Rank监督

arXiv:2610.33318v1 Announce Type: cross Abstract: Large language model agents launch simulations on high-performance computing (HPC) systems, where one sbatch or mpirun can commit thousands of node-hours and a forged result can cost lives. Of 23 surveyed systems, none supervises …

本文如何被排名

报道来源 [1]

面向MPI和批处理调度HPC系统的LLM代理确定性安全运行时：共识治理、硬件信任的每个Rank监督

相关实体

相关话题