Indian Institute of Management
| Name | Entity Origin | Action Guarded | PCTL Rule | Status | Last Modified | Actions |
|---|
Use this portal to configure the Deterministic Guardrails (Phase 3). The Azure Fabric will compile these rules into `.prism` format and deploy them to the isolated ANSS Function App (`anss-dtmc-validator-func`).
Instead of traditional RBAC, ANSS mathematically models AI execution as a Markov Chain. Defining a constraint guarantees zero probability of an AI agent reaching an illegal state, regardless of prompt injection attacks.
Describe your security policy in natural language, and the Meta-Agent will synthesize the formal PRISM state-space constraints.
Generate an `X-Intent-Manifest` to prevent Confused Deputy attacks. The middleware will strictly bind downstream tool execution to the capabilities signed in this token.
The semantic egress router constantly models outgoing text generation as a Discrete-Time Markov Chain (DTMC). Adjust the probabilistic threshold to balance sensitivity against "Stochastic Data Leaks".
Lower values aggressively terminate streams if generic data leakage patterns are detected.
Formal Verification Logic: Synthesized Markov Decision Process for symbolic safety checks.
Upload trusted knowledge artifacts (PDF, MD, txt) that the ANSS Watcher will use as verified ground-truth during contextual mediation.