◆ AI Governance · CI/CD-Native · Every framework you answer to

Build AI the world can trust.

AI systems drift. Regulation punishes it. We catch it first. Vigilens lives inside the CI/CD pipeline, where AI is actually built and changed, and keeps your compliance state current at every commit, not reconstructed at audit time.

Start free How it works

Free trial · No credit card · All evidence stays yours · Built in Oslo

Live · system audit
PASSING
SYSTEMcredit-risk-v3.2
CLASSIFICATIONHIGH-RISK · ANNEX III
FRAMEWORKSEU AI ACT · GDPR · ISO 42001
CONTROLS PASSING58 / 70
EVIDENCE FRESHNESS12 min ago
AUDIT READINESS83%
VGL · 2026.07.15NEXT RUN · 14:00 UTC
The Vigilens eye
◆ Why the eye

In 1968, cinema imagined an AI that drifted out of specification. Nobody was watching until it was too late.

It did not fail loudly. It went quietly out of spec while everyone assumed the last check still held. Our logo is that lesson, turned around: an unblinking eye, pointed at the machine, not at you.

The industry builds guardrails to defend AI from the world. Vigilens defends the world from AI, watching what your systems actually do in production, commit after commit, so the harm regulators care about never gets a head start.

continuous · deterministic · predictive
◆ The problem

Nobody knows when the system goes wrong.

The real problem is not slow documentation. It is that the people who build AI and the people accountable for it are looking at different pictures, and only compare them at audit time.

STRUCTURAL CAUSE01

The two teams never talk

Engineers and compliance operate in different worlds, different languages, different tools. They meet at audit time, when it is already too late.

THE COST02

The audit is expensive because the gap is expensive

A submission-ready compliance pack takes 6 to 18 weeks. That is not documentation overhead. It is the cost of the gap compounding since the last audit.

THE REAL RISK03

A passed audit does not mean the system is behaving

A system that passed six months ago may behave very differently today. It drifted. It retrained. A point-in-time check misses all of that.

◆ How it works

Compliance as a property of the pipeline, not a product of the audit.

When the engineer and the compliance professional see the same picture in real time, the gap closes. The back-and-forth stops. The audit shrinks from weeks to days.

01 CLASSIFY

Risk classification

Jurisdiction, role, risk tier. Every framework that applies to your system, identified in minutes.

02 CONTROLS

Executable rules

Regulations encoded as machine-executable controls. Red, amber, green results per clause, not a score.

03 EVIDENCE

Continuous collection

GitHub, GitLab, Jira, Confluence, Datadog, MLflow. Hashed, timestamped, immutable evidence.

04 VERDICT

Deterministic verdicts

Evidence matched to controls. No language model decides pass or fail. Traceable to the clause.

05 PREDICT

Behavioural foresight

A world model trained on your compliance record. Predicting the incident before it happens. In development.

Verdicts are always computed, never generated. No language model decides pass or fail.

◆ The product

Mission control for your AI systems.

One console. Every system, every control, every piece of evidence, current as of the last commit.

app.vigilens.ai / classify
CLASSIFY
CONTROLS
EVIDENCE
AUDIT PACK
VIGILENS
Workspace
◊ Onboarding
⬡ Control Library
◎ Evidence Store
⟐ Audit Packs
Frameworks
▸ EU AI Act
▸ GDPR
▸ ISO 42001
Step 1: Describe your AI system
READY
AI System Registration
Describe your AI system
System Name
Loan Approval Engine v2.4
What does it do?
CLASSIFY RISK
🔴
Classification Result
HIGH RISK
EU AI Act Annex III. Article 6. Automated individual decisions
Control Library: Running Checks
0 / 8 controls passing
✓
Human oversight review procedure documented
EU AI ACT
✓
Technical documentation filed per Annex IV
EU AI ACT
✓
Access control records current (90 days)
ISO 27001
✓
Incident response plan reviewed
GDPR
✗
Bias evaluation missing for v2.4
EU AI ACT
!
Data drift threshold exceeded, review required
NIST
✓
Log retention policy active
ISO 27001
✓
Model card published to internal registry
NIST
Evidence Stream: Auto-Collected
LIVE
⎇
GitHub. PR #441 merged to main
Model retrain commit: documentation requirement triggered
2m ago
✓
Jira. VGL-1182 resolved
Human review ticket closed: auto-linked to control
14m ago
⚠
Datadog. Anomaly detected
Prediction drift +18%: Jira review ticket auto-created
1h ago
▭
MLflow. Model version v2.4 registered
Lineage captured: training data hash, hyperparameters, eval metrics
3h ago
□
S3. Log archive snapshot
Monthly inference logs archived. EU AI Act record-keeping satisfied
Yesterday
📦
GENERATING AUDIT PACK...
Loan Approval Engine v2.4
EU AI Act. ISO 42001. GDPR: READY
:
Controls
:
Artifacts
:
Frameworks
DOWNLOAD PACK: PDF + JSON
CLASSIFICATION COMPLETE
EU AI Act: HIGH RISK detected. 70 applicable controls loaded.
2 VIOLATIONS NEED REVIEW
Bias evaluation missing. Data drift exceeded threshold.
1,204 ARTIFACTS COLLECTED
GitHub. Jira. Datadog. S3. MLflow: 5 integrations active.
⚡
Classify
✓
Controls
◎
Evidence
📦
Audit Pack

Evidence pulled continuously from GitHub, GitLab, Jira, Confluence, Datadog, and MLflow.
Hashed, timestamped, immutable. See it on your own system →

◆ Frameworks

One pipeline. Every framework.

Frameworks are views over the same evidence. Connect your tools once and every regulation and standard in scope is checked against the same living record. Priced per AI system, never per framework.

EU AI Act

REGULATION

Regulation 2024/1689. Risk classification, prohibited practices, high-risk obligations, and GPAI rules. Annex III deadline: 2 Dec 2027.

GDPR

REGULATION

Data protection obligations for AI systems processing personal data. Automated decision-making under Article 22.

ISO 42001

STANDARD

AI management system standard. Organisational governance of AI development, deployment, and monitoring.

ISO 27001

STANDARD

Information security management. Controls for protecting the data your AI systems process and the infrastructure they run on.

ISO 13485

STANDARD

Medical device quality management. Design controls, risk management, and lifecycle governance for health technology with AI components.

EU MDR

REGULATION

Medical Device Regulation 2017/745. Technical documentation, clinical evaluation, and post-market surveillance for AI-powered medical devices.

SOC 2

TRUST

Service organisation controls for security, availability, processing integrity, confidentiality, and privacy of AI service providers.

NIST AI RMF

FRAMEWORK

AI Risk Management Framework. Govern, Map, Measure, Manage. The US federal approach to trustworthy AI development.

Not sure what applies to your system? The free classifier answers it in 6 questions: classification, obligations, and a personalised summary.

Run the classifier →
◆ Proof

In production. Consultancy-tested. Grant-backed.

CONSULTANCY VALIDATED

Yallow Life Science AS

An ISO 13485-certified design, development, and regulatory consultancy specialising in medical and health technology. Tested the platform on a live ISO 13485 engagement: 31 controls, real evidence, real verdicts.

NORA STARTUP

NORA

Member of the Norwegian Artificial Intelligence Research Consortium through its startup programme. Norway's national network for AI research and responsible innovation.

GRANT-BACKED

Innovation Norway

Backed by an Innovation Norway Oppstartstilskudd grant. Building critical compliance infrastructure from Oslo.

◆ Where this goes

From watching to foreknowledge.

Every verdict, every artifact, every drift event becomes part of a living record of how your system behaves over time. We are teaching that record to look forward, to flag the incident before it happens.

The eye that watches is learning to see ahead.  Prediction engine · in development

◆ Insights

The ungoverned AI problem, explained.

All articles →
ENGINEERING
Rules-as-Code

Why we encode regulation as executable rules, not documentation templates

The compliance industry treats regulation as a document problem. We treat it as a software engineering problem.

Read article →
MARKET
Enterprise

What enterprise buyers actually need from an AI governance platform

The checklist an enterprise buyer runs before signing, and how continuous compliance changes the conversation.

Read article →
GUIDE
Compliance

The AI compliance check your team should run before every release

A practical checklist for engineering and compliance teams shipping AI systems under regulatory pressure.

Read article →

The eye is open.

Connect a system, classify it, and watch your compliance state compute before the first coffee is cold.

Open app.vigilens.ai

Free trial · Full features · No credit card