Govern What
AI Decides
Inventories record which AI you run. Prometheux governs what it decides: your policies run as executable rules over your data, so every answer follows the logic you approved and carries the evidence to show a regulator how.
Proven in production
Prompts Suggest.
Rules Enforce.
Guardrails filter what a model says, and registries record what you intended. When the decision matters, the policy needs to run as code, on the data, every time, and leave evidence behind.
Your Policies,
Executed Exactly
Prometheux encodes business policy as an ontology: readable rules over your own sources. Agents and apps call those rules instead of improvising them, so the parts of AI you can't leave to chance are deterministic.
Rules Before Models
Eligibility, limits and exposure logic run in the reasoning engine. The LLM explains the result; it doesn't decide it.
Defensible Decisions
Reasoning provenance and explanations show which rules applied to which records, ready for audit or review.
Agreed Definitions as Context
Pin approved policy documents and glossaries as static context, so LLM steps work from the same definitions as everyone else.
Instructions Kept Apart From Data
LLM concepts frame substituted data separately from instructions and neutralise injection markers, as a mitigation alongside your rules.
Put AI into regulated workflows with controls you can prove. Change a policy once and every agent follows it. Answer the regulator with rules and rows, not screenshots.
Governance You
Already Have
Prometheux runs inside your existing environments, as SaaS, in your VPC or on-premises, and inherits your governance instead of redefining it. AI gets access to governed knowledge, not a fresh copy of your data to secure all over again.
Data Stays in Place
Reason over warehouses, lakes and databases where they are. No copies to secure, sync or explain to compliance.
Platform-Native Access Control
Your existing authentication, access controls and security boundaries continue to apply.
Credentials Under Control
LLM keys live in saved connections. Tokens are masked by default, with OAuth 2.0 M2M and PAT support.
Assistant With Brakes
The AI Assistant works with your own access rights and asks before making changes and always before deleting anything.
Add AI without adding a new governance silo. Keep data sovereignty and your security model intact.
From Policy Text
to Governed AI
Connect in Place
Point Prometheux at your sources. Access stays governed by the controls you already run.
Write the Policy as Rules
Turn the policy into Vadalog rules: readable by risk, compliance and engineering.
Route AI Through the Rules
Agents and apps call those rules via APIs or MCP, so decisions come from the rules, not the model.
Show Your Work
Every result carries its provenance, and explanations turn it into a reviewable narrative.
Rules That Hold
Under Scrutiny
Explainable Credit Decisions
Banca Sella explains every credit decision with rules over the client's full history, cutting a regulatory flow from 5 hours to under a minute.
Read the case study →Traceable Advice
Recommendations follow encoded suitability rules, so each one can be traced and explained when a regulator asks.
Agents on a Rulebook
Agents call governed concepts through APIs or MCP instead of writing their own logic against raw tables.
Power AI →One Policy, Every Channel
A rule changed once flows to every report, app and agent built on those rules, with no drift between copies.
“Generative AI without ontologies is an engine without a road. Prometheux builds the road.”
Marco RomeiDeputy CIO · Banca Sella
“Prometheux enabled us to build a system that can access our databases, process all the client's historical variables, and provide an intelligent, understandable explanation autonomously.”
Massimiliano MalinHead of Credit Risk · Banca Sella
Questions,
Answered
What is AI governance in Prometheux?
It means the rules that constrain AI are written as executable code, run deterministically over your data, and leave a record of how each result was derived. Policies stop being prompt instructions and become logic the engine enforces.
Is this an AI governance platform for inventories and risk assessments?
No, and it works alongside one. Registries and risk platforms track which AI systems you run and how they were assessed. Prometheux governs what AI does at decision time: the policy runs as rules over your data, and each result records which rules produced it.
Does it map our controls to the EU AI Act or NIST?
Prometheux doesn't ship regulatory policy packs. It supplies what those frameworks ask you to show at decision level: the policy running as rules, and for each result a record of which rules and rows produced it. Use it alongside the platform that holds your AI inventory and assessments.
Does Prometheux replace our access controls?
No. Prometheux runs inside your environments and inherits your existing governance, authentication and security boundaries. Data stays in place.
Can we stop an LLM from ignoring a policy?
Decisions that matter are derived by the reasoning engine from your rules, not by the LLM. Agents call those rules through APIs or MCP instead of writing their own logic, the AI Assistant asks before making changes and always before deleting anything, and LLM steps keep instructions and data apart, with their outputs treated as model-generated content.
How do we show auditors what happened?
Chase provenance records each derived fact with the facts and rule that produced it, and explanations turn that into readable text.
Related Capabilities
Turn Your Policies
Into Running Code
Bring one policy and the data it applies to. We'll show it running as governed, explainable rules.
