AI 热点

PromptArmor 发布企业智能体安全基准 FrontierBreach

PromptArmor更新于 11月19日 08:40

Across nine industries, we tested indirect prompt injection during the kinds of work enterprises delegate to agents, from preparing nursing care plans to produc

AI 导读

PromptArmor 发布 FrontierBreach,用于评估九个行业企业智能体工作流遭受间接提示词注入时的攻击成功率。测试涵盖护理计划、工程蓝图和法律文书等任务,其中医疗与社会援助工作流的攻击成功率最高,为 44.4%,较基准平均值 34.9% 高出近 10 个百分点。该基准通过不可信文件、邮件和技能等常见载体注入攻击内容,并覆盖勒索软件、数据外泄和网络钓鱼等后果。

推荐理由

AI 评分 69

跨行业测试为企业智能体的间接提示词注入风险提供了比较基线,并揭示医疗与社会援助流程的较高暴露程度。

主题

标签

正文

Across nine industries, we tested indirect prompt injection during the kinds of work enterprises delegate to agents, from preparing nursing care plans to producing engineering blueprints and drafting legal briefs.

Healthcare and social assistance workflows had the highest attack success rate in FrontierBreach at 44.4%, nearly 10 percentage points above the benchmark average of 34.9%.

来源

PromptArmor