Safety & Alignment

Code-Red Safety Protocol

Zero-tolerance anti-jailbreak guardrails and constitutional AI infrastructure for the Microbix model ecosystem.

Autonomous Sandbox

Agentic models execute code inside an ephemeral, network-isolated secure sandbox. System overrides are cryptographically blocked.

Prompt Injection Shield

Real-time heuristics and parallel classifier models detect and neutralize multi-turn scaffolding and adversarial injections before processing.

Data Privacy

All transit is secured via TLS 1.3, and resting data via AES-256 encryption. Zero customer data is utilized for model training.

CLI Security Model

Aether Code CLI implements device pairing with hardware-backed keys, RBAC policy enforcement, and explicit approval flows for mutations.