PlatformThe engine behind Governance AI

Every capability
behind Governance AI

An index, not a pitch. Governance AI is built from all seven. SOC is deliberately separate — it runs its own pipeline for code and infrastructure, not AI traffic. Listed in the order a request meets them: inspected, enforced, constrained, recorded — then tested, measured, and reported against the frameworks your auditors use. Each has its own page.

Capabilities
7one platform
Guard types
13shipped
LLM providers
40+ supported
Frameworks mapped
5self-assessed
01 / Index

Inspect, enforce, record —
then prove it.

The first four capabilities act on traffic in real time. The last three answer the question every auditor eventually asks: how do you know it worked?
01 — INSPECT

Inspection

Synchronous, in-path inspection of prompts and responses, with rules scoped to tool-call arguments. Streamed responses are checked after delivery. 13 guard types (8 model-backed, 5 rule-based) plus keyword, regex, and LLM-judge criteria.

13 guard types · in-path
Read more →
02 — ENFORCE

Policy

Policy Sets group Criteria and bind them to AI Agents with an action — block, flag, redact, or human review. Versioned, with a draft → pending review → approved → published lifecycle.

Policy Sets · per-agent
Read more →
03 — CONSTRAIN

Allow-lists

Bind each agent to an approved provider and a list of approved models. A call to another model is flagged, or blocked if you turn blocking on, with allowed models per use case, and access keys scoped to specific agents.

Providers · models · keys
Read more →
04 — RECORD

Audit ledger

A timestamped record of requests, policy decisions, configuration changes, and human review decisions. Export it to CSV or JSON for an auditor.

Logged · CSV + JSON export
Read more →
05 — TEST

Evaluations

Benchmark and red-team registered agents against the Project Moonshot catalogue from AI Verify Foundation — safety, bias, jailbreak resistance, privacy leakage, and capability, graded per recipe.

Project Moonshot · graded
Read more →
06 — MEASURE

Ethics & bias

A nightly job samples a share of the previous day’s traces and scores them against six LLM judge criteria — toxicity, demographic bias, sentiment consistency, fairness, misinformation, and privacy leakage.

6 judges · nightly
Read more →
07 — REPORT

Framework mapping

Coverage against NIST AI RMF, ISO/IEC 42001, the OWASP LLM Top 10, Malaysia’s AISCF, and Singapore’s MGF for GenAI — computed from your own activity, with the evidence behind each category.

5 frameworks · CSV export
Read more →
08 — TOGETHER

Governance AI

All seven, working as one product: a live dashboard, a violations pipeline with human review, a registry of every AI Use Case and Agent, and the evidence trail underneath it.

7 of 7 · one ledger
See Governance AI →
09 — A SEPARATE PIPELINE

SOC

Logs, metrics, traces, CVE Radar, and threat intel run on their own pipeline — built for code and infrastructure, not AI traffic. It keeps its own audit trail rather than sharing the ledger above.

0 of 7 · own audit trail
See SOC →