- Core
- AI Act
AI System Audit
Full examination of an AI system in production: data, model, integrations, monitoring. Risk, robustness and compliance in a single dossier.
For whom Those who already have a live AI system and must prove it holds up.
EU Regulation 2024/1689 · independent audit of AI systems
We look beneath the surface of AI systems. Black on white.
We inspect your AI systems properly — risk, security, transparency, bias. When everything is in order we attest it, with documentation ready for authorities, clients and suppliers. When it isn't, we give you a prioritized remediation plan and a date.
AI Act deadlinesLIVE
- Literacy obligation02.02.2025
- Penalties active02.08.2025
- Inspections start02.08.2026
EU AI ACT · ENFORCEMENT FROM 02.08.2026
- 001 · Risk
The right class
We classify every AI system by the AI Act's risk category — prohibited, high, limited, minimal — and measure its real exposure, not the declared one.
4+AI Act categories
- 002 · Transparency
Explainable & traceable
We verify model explainability, data provenance and quality, logging and the technical documentation the Regulation requires.
100%Data chain
- 003 · Compliance
Gap, then remediation
EU AI Act + GDPR gap assessment with a prioritized remediation plan, owners and deadlines. Documented and attested, black on white.
0Surprises at inspection
Audit services
Six ways to look underneath.
From a single model to your entire supplier base. Every audit closes with a navigable report and a remediation checklist — not a PDF that gathers dust.
- AI Act
Vendor & Supply Audit
We evaluate the AI vendors and third-party models you use: contractual clauses, liability, upstream compliance. The risk you inherit, mapped.
For whom Those who integrate third-party AI and must answer for it downstream.
- Most requested
- AI Act
EU AI Act Gap Assessment
Where you stand against the Regulation and what's missing to get there. Distance measured, actions prioritized, owners and dates. The compass before inspections.
For whom Those who don't yet know where to start with compliance.
- Specialist
Bias & Fairness Review
Statistical tests on datasets and outputs: disparities between groups, hidden proxies, drift. Documented, repeatable fairness metrics.
For whom HR-tech, credit, scoring and every decision that touches people.
- Specialist
Security & Robustness
Prompt injection, data leakage, adversarial input, hallucinations in critical contexts. We stress the system before others do.
For whom Those who expose AI to end users or sensitive data.
- Core
Documentation & Attestation
We put the technical documentation in order and produce the audit attestation: the proof, dated and signed, to show authorities, clients and tenders.
For whom Those who have fixed their systems and must prove it to third parties.
Every audit includes a debrief call and 30 days of follow-up.On-site nationwide · remote across the EU
The method
Scope, Inspect, Report, Attest.
An audit isn't an opinion: it's a repeatable protocol. The same rigor for a chatbot and for a scoring system.
- PHASE 01
Scope
AI systems inventory, risk class, audit perimeter and evaluation criteria agreed. You know exactly what we'll look at.
Perimeter + criteria
- PHASE 02
Inspect
Technical and documentary examination: data, model, logging, security, bias, governance. Evidence collected, not impressions.
Evidence · tests
- PHASE 03
Report
Navigable report with findings classified by severity, gaps against the AI Act and a remediation plan with owners and dates.
Report · plan
- PHASE 04
Attest
Once gaps are closed, we issue the audit attestation and compliance documentation. 30-day recheck included.
Attestation · follow-up
FAQ
Everything you need to know.
The essential answers on what an AI audit is, what it produces and when you need it. If you can't find yours, write to us.
What exactly is an AI audit?
It's an independent, structured examination of an artificial-intelligence system: data, model, integrations, security, bias and documentation. We verify that the system does what it claims, within the AI Act's risk limits, and produce verifiable evidence.
It's not generic consulting: it's a protocol with declared criteria, classified findings and a measurable remediation plan.
Does the AI Act require auditing systems?
For high-risk systems, Regulation (EU) 2024/1689 requires risk management, data quality, technical documentation, logging and human oversight. Enforcement for Annex III high-risk systems starts on 2 December 2027. An audit is the most direct way to prove these requirements are met.
How long does it take and what do you deliver?
From 1 to 4 weeks depending on scope. We deliver a navigable report (30-50 pages) with findings by severity, gaps against the AI Act, a remediation plan with owners and dates, and — once gaps are closed — a signed audit attestation.
What's the difference from training?
The audit looks at systems; training at people (art. 4 obligation). They're complementary: an audit often surfaces training needs, and a training path closes the human side of compliance.
Do you also audit third-party suppliers and models?
Yes. With the Vendor & Supply Audit (A2) we evaluate the third-party models and AI vendors you integrate: clauses, liability and upstream compliance. The risk you inherit becomes visible and manageable.
Does the attestation have legal value?
It's an independent professional attestation: it documents method, criteria and evidence as of the audit date. It doesn't replace a notified-body certification where required, but it holds up before authorities, clients and tenders as proof of due diligence.
Where's the best place to start?
With the Gap Assessment (A3): in 2-3 weeks you know exactly where you stand against the AI Act and what to do, in priority order. It's the safest way not to waste the first euro in the wrong place.
Request an audit
Let's put your systems to the test.
A 30-minute call to define the scope and risk class. You leave with a clear quote and a date — not a generic PDF.