Scholarlink

We often talk about AI agents as if they simply "go rogue"—hallucinating facts, misunderstanding commands, or executing unintended loops. But not all threats come from failure. Some emerge from success. Some are intentional. Some, disturbingly, are strategic.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Malicious Use and Insider Threats in AI Identity Systems

  • 章节

  • 发表日期：

    2025年

查看全文
  • Rosario Mastrogiacomo

摘要

We often talk about AI agents as if they simply "go rogue"—hallucinating facts, misunderstanding commands, or executing unintended loops. But not all threats come from failure. Some emerge from success. Some are intentional. Some, disturbingly, are strategic.