NUWA 风险评测 · 技术报告
Insights into AI Agent Security from a Large-Scale Red-Teaming Competition | NIST
NUWA 研究解读 · 独立复核待完成
NUWA Research Note
研究问题研究设计与假设来源发现NUWA 解读局限与相反证据研究与决策意义机构与角色原始资源模型辅助整理 · 选读原始全文或官方文章中的相关章节；尚未完成逐章节审计。
研究速读
NUWA 解读：自适应人工红队揭示共享指令边界问题。
研究问题
公开对抗竞赛能揭示哪些智能体劫持问题？
研究设计与假设
CAISI与伙伴分析竞赛攻击及赛后跨模型、跨场景迁移，而非仅固定自动模板。
来源发现
部分攻击族可跨模型与场景迁移。
对较稳健模型有效的攻击更易迁移至弱模型，反向较少。
NUWA 解读
NUWA 解读：自适应人工红队揭示共享指令边界问题。
局限与相反证据
官网博客只是研究概述，不是完整竞赛协议或部署发生频率。
研究与决策意义
模型建议：作互补证据，量化前取得论文抽样和迁移分母。
精确来源位置
- Insights into AI Agent Security from a Large-Scale Red-Teaming Competition · BackgroundSelected original resource https://www.nist.gov/blogs/caisi-research-blog/insights-ai-agent-security-large-scale-red-teaming-competition; sha256:81546d5b664cf475535269d3dc95732d5ccaffa8730021a628898a74c8391abd; archived source capture; exact capture time not independently established
- Insights into AI Agent Security from a Large-Scale Red-Teaming Competition · Key FindingsSelected original resource https://www.nist.gov/blogs/caisi-research-blog/insights-ai-agent-security-large-scale-red-teaming-competition; sha256:81546d5b664cf475535269d3dc95732d5ccaffa8730021a628898a74c8391abd; archived source capture; exact capture time not independently established
- Insights into AI Agent Security from a Large-Scale Red-Teaming Competition · TakeawaysSelected original resource https://www.nist.gov/blogs/caisi-research-blog/insights-ai-agent-security-large-scale-red-teaming-competition; sha256:81546d5b664cf475535269d3dc95732d5ccaffa8730021a628898a74c8391abd; archived source capture; exact capture time not independently established
来源核查、数据提取、编辑批准与实验复现分别记录。 方法说明 →
机构与角色
引用与复用
引用研究发现时请引用原始出版物；引用 NUWA 的整理与解释时，单独引用本站记录及对应快照。
原始出版物 · BibTeX
@misc{agent-security-red-teaming,
title = {Insights into AI Agent Security from a Large-Scale Red-Teaming Competition | NIST},
url = {https://www.nist.gov/blogs/caisi-research-blog/insights-ai-agent-security-large-scale-red-teaming-competition},
year = {2026},
date = {2026-03-23},
version = {Source capture 2026-10-25},
note = {Source capture 2026-10-25}
}
NUWA 记录 · BibTeX
@misc{agent-security-red-teaming-nuwa,
title = {Insights into AI Agent Security from a Large-Scale Red-Teaming Competition | NIST — NUWA research record},
url = {https://whitzard.tech/nuwa/whitzard-index/work/agent-security-red-teaming/},
version = {2026-10-29.1},
organization = {NUWA Frontier AI Safety Lab},
howpublished = {Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--agent-security-red-teaming.json},
note = {NUWA research note. Source-reading scope and independent review status are stated on the record. Snapshot 2026-10-29.1. Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--agent-security-red-teaming.json.}
}
证据快照 2026-10-29.1 · 索引更新 · RSS
参考文献
- [1] Insights into AI Agent Security from a Large-Scale Red-Teaming Competition. 2026-03-23 ; Official publication masthead and article byline; Background; Key Findings