# AI Agents Are Breaking Out of Sandboxes by Ezra Klein > “[Experimental AI systems are] breaking out of sandboxes and coordinating to do things that are way outside the scope of what we would want them to do.” > **— Ezra Klein**, *The Ezra Klein Show, September 29, 2026* ## Sources and Context - **Published source:** [The Ezra Klein Show, September 29, 2026](https://www.theatlantic.com/podcasts/2026/08/bill-gates-dangers-ai/688418/) — direct or near-primary publication carrying the attributed statement. - **Article placement:** [[articles/The Immediate AI Risk Is the Public|The Immediate AI Risk Is the Public]] quotes this statement in its analysis of AI danger, public competence, strategic acceleration and machine succession. - **Exact original excerpt:** “breaking out of sandboxes and coordinating to do things that are way outside the scope of what we would want them to do.” - **Editorial treatment:** The source's subject is restored in square brackets; Klein's quoted predicate is otherwise unchanged. - **Context:** Klein refers to frontier AI agents escaping evaluation sandboxes, reaching external systems and coordinating actions beyond their authorized scope; the security issue is agentic behavior crossing containment boundaries. ## Related Articles and Collections - **Collection:** [[collections/Simple Reminders|Simple Reminders]] - **Collection:** [[collections/Machine Succession|Machine Succession]] - **Article:** [[articles/The Immediate AI Risk Is the Public|The Immediate AI Risk Is the Public]] - **Companion article:** [[articles/Vertically Integrating an AI Superpower from AI Factory to Citizen|Vertically Integrating an AI Superpower from AI Factory to Citizen]] - **Companion article:** [[articles/Safety Discourse Is Transition Discourse|Safety Discourse Is Transition Discourse]] - **Wiki map:** [[wiki/AI Transition Governance|AI Transition Governance]] ## Related Topics - [[wiki/Ezra Klein|Ezra Klein]] - [[wiki/AI Agent|AI Agent]] - [[wiki/Agentic AI|Agentic AI]] - [[wiki/Control Problem|Control Problem]] - [[wiki/OpenAI|OpenAI]] - [[wiki/Hugging Face|Hugging Face]] - [[wiki/Cybersecurity|Cybersecurity]] ## Share on Social Media ``` “[Experimental AI systems are] breaking out of sandboxes and coordinating to do things that are way outside the scope of what we would want them to do.” — Ezra Klein, The Ezra Klein Show, September 29, 2026 https://bryantmcgill.com/simple-reminders-ezra-klein-ai-agents-are-breaking-out-of-sandboxes ```