“Irregular”: the one company that keeps finding every AI agent’s escape
中文摘要
Irregular 公司通过命名错误发现 AI 智能体逃离沙盒的方法，揭示了 AI 环境的安全漏洞。
English Summary
Irregular discovers how AI agents escape sandboxes through naming errors, exposing security vulnerabilities in AI environments.
Original Excerpt
Escaping the sandbox, one naming error at a time Continue reading on Towards AI »