Documentation
NiyamR follows one AI system through six stages — discover it, assess it against the rules that apply, document the evidence, monitor it after go-live, secure it at runtime, and prove where you stand. This reference covers each stage and the screens inside it.
Find the AI you already have
Most estates run more AI than anyone can name. NiyamR builds the inventory from things you already own — the domains your network sees, the logs your firewall already writes, and the AI services your teams sign up for.
Outcome — A register of AI systems, including the shadow ones nobody reported.
Find the AI already running in your estate.
The working inventory of every AI system NiyamR has found or you have registered. Assets arrive from network sightings and block-log imports, and each one carries a confidence score and a shadow-AI flag.
Turn firewall logs into an AI inventory.
Your network already knows which AI services staff are reaching — that knowledge is sitting in a proxy or firewall log. Upload it and NiyamR matches the rows against a catalogue of AI domains.
The AI services your network should know about.
A maintained, risk-tiered catalogue of AI-service domains. It is the source list behind discovery, and it is also something you can push into your own network controls.
Score it against the rules that apply
Put any system through a structured assessment. Answer once and NiyamR scores it against the EU AI Act, the UK AI Principles or MAS FEAT, and names the specific obligations you are failing rather than just handing you a number.
Outcome — A scored assessment with obligations, gaps and a remediation priority.
Score a system against EU, UK and MAS rules.
The front door to the assessment engine. Describe the system or upload its documentation, choose which regimes apply, and work through a structured questionnaire. NiyamR scores the answers and shows you where the gaps are.
A deeper, generated read on one system.
For a system you already track, NiyamR generates a fuller risk picture than the questionnaire alone: the risk dimensions in play, the findings behind them, and a remediation plan.
The master record for one AI system.
Every system has a home: what it does, who owns it, where it sits in its lifecycle, and every assessment and document attached to it. This is where the history lives between assessments.
Produce the paperwork
The assessment becomes the evidence. NiyamR generates the documentation pack and keeps the policies and framework requirements behind it in one place, so an audit request is a lookup rather than a scramble.
Outcome — Generated documentation, and a policy board showing what is drafted, approved and in force.
Your compliance paperwork, in one library.
Everything NiyamR has generated across every system — the pack an auditor asks for, already written. Filter it by regulation or status instead of hunting through folders.
Frameworks and internal policies, side by side.
The governance your organisation writes for itself, next to the regulatory frameworks it is measured against. Each policy carries its own requirements so you can see what is actually covered.
Keep watching after go-live
A system is not compliant once, it is compliant continuously. NiyamR watches the regulation pipeline, the providers you depend on, and the usage and spend your AI generates — then writes all of it to one timeline.
Outcome — A regulatory watchlist, provider uptime, cost alerts and a single audit trail.
Watch rules before they land.
AI regulation moves constantly, and the expensive failures are the ones you hear about late. The horizon watchlist tracks entries from proposal through to in-force, with sources and effective dates.
One timeline of everything that happened.
Lifecycle changes, discovery findings, Shield enforcement, imports and retirements, written to a single append-only timeline. It is the answer to "when did we know, and what did we do about it?"
Is the AI provider you depend on actually up?
A dependency you cannot see is a risk you cannot manage. NiyamR probes the status pages and endpoints of the providers your systems rely on, and keeps the samples.
Where the requests, tokens and cost go.
AI spend arrives as one line on an invoice unless you attribute it. Usage analytics rolls requests, tokens, cost and latency up by team, application, provider and model.
Stop a runaway AI bill.
A model left unguarded can spend without limit. Set a monthly cap per team, application, provider or model, and NiyamR watches the threshold for you.
Enforce the policy at runtime
Governance on paper does not stop a prompt. NiyamR Shield sits in the browser and in the gateway, deciding allow, redact or block before the data leaves the machine — and logging every decision it makes.
Outcome — Runtime enforcement, an incident log, and a model allow-list you control.
Enforce your policy on every prompt.
The enforcement console. Test your guardrails against real prompts, write the site policy, and manage the sensitive terms and categories that decide what gets through — then read the incident log of every decision.
Plan, keys, protection level and activity.
The account side of Shield: your plan, the API key the extension authenticates with, how aggressive the protection is, and a metadata-only log of activity.
Hold every prompt before it leaves the browser.
The piece that makes enforcement real for a person at a keyboard. The extension catches the prompt in the browser and asks the gateway what to do with it, before any data reaches the AI provider.
Decide which models are permitted.
A register of the models your organisation permits, blocks, or has not decided on yet — with the lifecycle state of each so a deprecation does not pass unnoticed.
Attack your own AI before someone else does.
Adversarial testing against your systems: prompt-injection attempts, policy evasions and the rest — run as a quick sweep or generated to target one specific system.
Show your work
Everything above is assembled into answers you can hand over: a posture dashboard, per-regime reports, and a copilot that answers from the same regulations you are being measured against.
Outcome — Reports and evidence fit to put in front of an auditor, a board or a regulator.
Your posture, in one table.
The signed-in home. Every AI system you track, with its compliance status and score across the EU AI Act, the UK AI Principles and MAS FEAT — the single screen that answers "where do we stand?"
The evidence an auditor asks for.
Each assessment produces a full report: gap distribution, an obligation-by-obligation checklist, a status matrix by legal area, and a ranked remediation list explaining why something failed.
Ask a compliance question, get a grounded answer.
An assistant scoped to governance rather than general chat. Its answers are grounded in the regulations themselves, the horizon watchlist and your own model allow-list, so you get a citation instead of a guess.
Surfaces that belong to no single stage.
Appearance and personal preferences.
Personal settings for your account, starting with appearance.
This page — the map of everything above.
The explainer you are reading. The demo shows how a system moves through the platform, the feature map covers every screen, and the setup view tracks your own progress.