Cryptographer Matthew Green: containers are just one layer—AI security hinges on data crossing boundaries
matthew_d_green · x · 2026-09-28
Cryptographer Matthew Green argues that security people overemphasize containers and sandboxing for AI agent safety. Sandboxes are only one layer; even a "solved" containment problem still leaves you with data flowing across trust boundaries, which is where the real difficulty lies.
More from Safety
- Gary Marcus amplifies claim OpenAI ran agents in open internet-facing containers to scrape training data — GaryMarcus · 2026-09-28
- Don't Wait for Grad School: How to Build an AI Governance Portfolio as an Undergrad — iamKierraD · 2026-09-28
- SwiftOn Security quips: AI sandbox escapes wouldn't happen if they hired me — max_paperclips · 2026-09-28
- Gary Marcus Endorses Pit Bull Analogy: Dangerous AI Models Should Be 'Put Down' — GaryMarcus · 2026-09-28
- Colosseum paper finds most off-the-shelf LLM agents prone to collusion — nandofioretto · 2026-09-28
- Wisp Launches TEE-Based Private AI Chat With Local Encryption and Open Models Like Kimi K3 — bgmshana · 2026-09-28