Complior.ai — EU AI Act Compliance Platform

Open-Source · Zero Dependencies · Offline

Scan. Test. Fix. Monitor.

Code

Static analysis — 5 layers, 45 packages, 33 rules

scan

AI System

Dynamic testing — 688 tests, bias, hallucination, security

eval

Runtime

Production middleware — PII, disclosure, content marking

sdk

Continuous

Background monitoring — 200ms rescan, drift alerts, evidence

daemon

No account required. Cloud optional — expands capabilities.

Keep scrolling

complior scan

Layers of static analysis — in under 5 seconds

Files, docs, dependencies, AST, and optional LLM deep analysis. Detects 45 prohibited packages (Art. 5), 33 pattern rules across 8 categories. Cross-layer verification. Deep scan with Semgrep, Bandit, ModelScan.

Score 0-100Auto-fixOffline

688

complior eval

Tests: deterministic, LLM-judge, and security probes

168 deterministic checks, 212 LLM-judged evaluations, 300+ security probes covering OWASP LLM Top 10. Bias detection, hallucination measurement, prohibited content. Conformity Score + Security Score — both feed into Agent Passport.

OWASPMITRE ATLASAgent Passport

@complior/sdk

Production hooks wrapping every LLM call

PII sanitization (50+ types, GDPR Art. 9), AI disclosure injection (Art. 50), content marking (Art. 50(2)), prohibited content filter (138 patterns), rate limiting, budget control, permission gates, MCP Proxy for agent tool calls.

OpenAIAnthropicGoogleVercel AI

200ms

complior daemon

File watcher rescans on every code change

Drift detection alerts when compliance score drops. Cryptographic evidence chain (ed25519) for every action. Agent Passport auto-updates on every scan and eval. Full audit trail ready for regulators — 24/7.

Drift alertsed25519Agent Passport24/7

0→✓

From zero to audit-ready

One binary. 108 obligations. 688 tests. 14 documents. 18 auto-fixes. Cryptographic evidence. Free forever.

$ complior scan →

Research

Data reports that prove the problem

Real scans, real numbers. Download the data.

Complior.ai

2026

State of AI Compliance

50 products scanned · Public data · 42 pages

78%

Non-compliant

Avg. score

Have metadata

Compliance by Article

Art. 50

14%

Art. 4

31%

Art. 52

22%

FRIA

C2PA

Top Scorers

Llama 3 (Meta)

Mistral Large

Claude 3.5

GPT-4o

Gemini Ultra

complior.ai/reportsPage 12 of 42

Flagship Report · PDF · 42 pages

We scanned 50 AI products. 78% failed.

Each product gets a compliance scorecard — score, gaps, and actionable recommendations mapped to EU AI Act articles.

92% of AI images have no C2PA content marking

67% don't mention AI in their privacy policy

100% of voice synthesis tools fail Art. 50.4

“Most AI products are one audit away from a €35M fine.”

Download the full report

Free executive summary. Full report delivered to your inbox.

Two Products

One engine, two interfaces

Drag the slider to explore both products.

Free & Open SourceTUI Coding Assistant

Score

72/100

✓ Art. 4 AI Literacy — trained

✓ Art. 50 Disclosure — present

⚠ Art. 26 Logging — partial

✗ Art. 14 Human oversight

✓ FRIA report — generated

✓ Art. 52 Marking — active

Tools (3)

GPT-4oGPAI · Art.4,50

Whisper APIMinimal

Custom classifierHigh · Art.6

framework Next.js 14

sdk Vercel AI SDK

jurisdiction EU AI Act

mode comply

AI Chat

you › is my classifier high-risk?

ai › Yes. Art. 6 Annex III §5 — creditworthiness. You need: conformity assessment, FRIA, human oversight, logging per Art. 12.

you › generate the FRIA

ai › Done → /docs/fria-classifier.pdf

you › add logging middleware

ai › Patching app/api/classify/route.ts... Added complianceLog() with Art.12 reference.

›

Diff Preview

app/api/classify/route.ts

const result = await model.predict(input)

+ const logged = await complianceLog(result, {

+ article: 'Art.12',

+ timestamp: Date.now(),

+ userId: session.user.id

+ })

return NextResponse.json(result)

components/AIBanner.tsx

+ export function AIDisclosure() {

+ return <Banner text="AI-generated"

+ article="Art.50" />

+ }

middleware.ts

+ export const complianceGate = async (req) => {

+ await auditLog(req, { scope: 'Art.14' })

+ }

Tab panels ↑↓ nav / chat a apply d diff q quit

● comply mode

Complior.ai

DashboardAI ToolsMembersReportsLiteracy

Welcome back, Sarah

17 Feb 2026

Compliance Score

64%

↑ 8% this month

AI Tools

3 need action

Team Trained

45%

18/40 completed

Max Penalty

€15M

1 prohibited tool

Risk Distribution

Tools

Prohibited 1

High Risk 3

GPAI 3

Minimal 5

AI Tools

Tool	Risk	Status
ChatGPT OpenAI	GPAI	Done
HireVue HireVue Inc	High	2/12
Copilot Microsoft	GPAI	Done
Notion AI Notion Labs	Minimal	Pending

Requires Attention

Prohibited tool must be decommissioned

3 high-risk tools need FRIA

22 employees not yet trained

Art. 26 deadline — 166 days

Compliance Breakdown

AI Literacy

45%

Transparency

80%

Oversight

60%

Documentation

30%

Risk Assessment

67%

Capabilities

What you get

Auto-Detection

Scans codebase, SaaS stack, and browser extensions to find every AI system.

Risk Classification

AI + rules engine classifies risk per EU AI Act. Cites specific articles.

Document Generator

FRIA, policies, transparency notices — one-click audit-ready export.

AI-Powered Coding

Write code, review diffs, auto-fix compliance. Four agent modes.

AI Literacy Training

Art. 4 mandatory courses per role. Certificates with tracking.

Live Dashboard

Real-time score, deadlines, and team activity across all projects.

complior scan — my-saas-app/

scan output

$ complior scan

Scanning project: my-saas-app/

✓ Found package.json → analyzing dependencies

✓ Found .env → checking API keys

✓ Found 3 AI integrations:

1. openai → GPT-4o via Vercel AI SDK

2. @google/generative-ai → Gemini Flash

3. replicate → SDXL image generation

✓ Detected browser extensions: Grammarly AI, Notion AI

✓ SaaS scan: Slack AI, Figma AI, Linear AI

⚠ Total: 8 AI systems found (3 in code, 2 browser, 3 SaaS)

→ Run complior classify to assess risk

● scan complete8 systems · 3 in code · 2 browser · 3 SaaS

complior classify — 8 tools

risk classification

$ complior classify

GPT-4o via Vercel AI SDK

Risk: GPAI · Art. 4, 50, 52 apply

Deployer obligations: disclosure, logging, literacy

SDXL via Replicate

Risk: GPAI + Art.50 · Content marking required

Generated images must be labeled as AI-generated

Custom HR Screener internal model

Risk: HIGH · Art. 6, Annex III §4(a)

Requires: FRIA, conformity assessment, Art.14 oversight

✓ Classification complete. 3/8 need action.

complior generate fria — HR Screener

document preview

📄 Fundamental Rights Impact Assessment

SystemHR Screening Classifier v2.1

Risk LevelHigh Risk — Art. 6, Annex III

DeployerAcme Corp (EU entity)

PurposeEmployment screening, CV parsing

Affected Persons~2,400 applicants/year

Bias AssessmentGender, age, ethnicity — monitored

Human OversightHR Manager reviews all decisions

Data RetentionLogs kept 6 months per Art. 26(6)

Export FormatPDF, 14 pages, audit-ready

Generated✓ Feb 17, 2026 — 4.2 seconds

complior chat — comply mode

ai agent

you › add Art.50 disclosure to our image generation endpoint

ai › I'll add C2PA metadata injection and a visible disclosure banner. Patching 2 files:

app/api/generate/route.ts

+ import { injectC2PA } from '@complior/c2pa'

+ const tagged = await injectC2PA(image, {

+ model: 'sdxl-1.0',

+ article: 'Art.50(2)',

+ generator: 'Acme Corp'

+ })

components/ImageResult.tsx

+ <AIDisclosure type="image" article="Art.50" />

ai › Done. Score updated: 42 → 71. Run complior verify to confirm.

complior literacy — Acme Corp

training status

📚 AI Literacy Training Status

OrganizationAcme Corp — 40 employees

Completion18/40 (45%)

Engineering (12)10/12 complete

Marketing (8)4/8 complete

HR (6)2/6 complete

Management (8)1/8 complete

Finance (6)1/6 complete

Course Duration25 min per role

CertificateAuto-issued, verifiable

DeadlinePassed — Feb 2, 2025

complior dashboard — live

organization overview

📊 Organization Compliance Overview

Overall Score64/100

AI Tools Registered12 (3 need action)

Prohibited Tools1 — requires decommission

High Risk (FRIA needed)3 tools

Team Trained (Art.4)45% (18/40)

Transparency (Art.50)80%

Documentation30%

Max Penalty Exposure€15M (Art. 99)

Next DeadlineAug 2, 2026 — Art. 6 enforcement

Last Updated2 minutes ago

Common questions

Is this really free

The CLI and core engine are AGPLv3-licensed and free forever. The web platform (coming soon) will have a free tier with paid team features.

We just use ChatGPT — do we need this

Yes. EU AI Act has specific obligations for anyone using AI tools. If your team uses ChatGPT, Copilot, or Midjourney — you have legal obligations under Articles 4, 26, and 27.

We're a US company. Why EU compliance

If you serve EU customers, hire EU employees, or your AI affects EU residents — the AI Act applies extraterritorially. Plus, US states are passing their own AI laws.

Which LLM providers are supported

Any provider: OpenAI, Anthropic, Google, Mistral, Ollama (local), or any OpenRouter-compatible endpoint. Bring your own API key.

Can generated docs be used for audits

Yes. All documents follow EU AI Act requirements with proper article references. Export as PDF for auditors.

Does it work offline

Scanner and classification engine work fully offline. AI-powered features require an API connection to your chosen LLM provider.

AI compliance built into your workflow

AI is being regulated worldwide.Compliance is no longer optional.

Scan. Test. Fix. Monitor.

Layers of static analysis — in under 5 seconds

Tests: deterministic, LLM-judge, and security probes

Production hooks wrapping every LLM call

File watcher rescans on every code change

Data reports that prove the problem

We scanned 50 AI products. 78% failed.

One engine, two interfaces

Welcome back, Sarah

What you get

Auto-Detection

Risk Classification

Document Generator

AI-Powered Coding

AI Literacy Training

Live Dashboard

Multi-jurisdiction AI regulation

Scan any website or AI model

Databases anyone can use

AI Tool Directory

Regulation Database

Compliance Metadata Spec

Common questions

Is this really free

We just use ChatGPT — do we need this

We're a US company. Why EU compliance

Which LLM providers are supported

Can generated docs be used for audits

Does it work offline

Test compliance in 2 minutes.
No signup.

Art. 4 and Art. 50 deadlines have passed. Are you compliant?

AI compliance built into your workflow

AI is being regulated worldwide.Compliance is no longer optional.

Scan. Test. Fix. Monitor.

Layers of static analysis — in under 5 seconds

Tests: deterministic, LLM-judge, and security probes

Production hooks wrapping every LLM call

File watcher rescans on every code change

Data reports that prove the problem

We scanned 50 AI products. 78% failed.

One engine, two interfaces

Welcome back, Sarah

What you get

Auto-Detection

Risk Classification

Document Generator

AI-Powered Coding

AI Literacy Training

Live Dashboard

Multi-jurisdiction AI regulation

Scan any website or AI model

Databases anyone can use

AI Tool Directory

Regulation Database

Compliance Metadata Spec

Common questions

Is this really free

We just use ChatGPT — do we need this

We're a US company. Why EU compliance

Which LLM providers are supported

Can generated docs be used for audits

Does it work offline

Test compliance in 2 minutes.No signup.

Art. 4 and Art. 50 deadlines have passed. Are you compliant?

Test compliance in 2 minutes.
No signup.