| Framework Home | Research Index | Cross-Pillar Governance | AISM | NEXUS | Dashboard |
Current framework context: AI SAFE² v3.1. This research note preserves its original publication date, evidence, and historical framework references. Use current v3.1 normative control and profile documents for implementation or conformance decisions.
ID: RN-2025-006 | Related Control: [P1.T2.1], [P1.T2.1_ADV] | Status: Verified
Container Breakouts: Standard Docker containers share the host kernel. If an Agent has code-execution capabilities (e.g., Python REPL), a kernel exploit allows it to escape the container and compromise the host infrastructure.
For Tier 3 (Agentic) systems, standard containerization is insufficient. We mandate User-Space Kernels.
All agents capable of writing code must run in gVisor (runsc) or AWS Firecracker microVMs. This provides a distinct kernel boundary, mitigating escape vulnerabilities.
Agents should not have persistent filesystems. Containers must be destroyed and rebuilt after every task execution to prevent malware persistence.
| Previous research note | Research Index | Next research note |
| Framework Home | Cross-Pillar Governance | NEXUS | Challenge Lab |
AI SAFE² v3.1 | Cyber Strategy Institute