What is the EU AI Act enforcement deadline?

August 2, 2026 is when EU AI Act obligations for high-risk AI systems become enforceable. Penalties for non-compliance can reach 7% of global annual revenue or €35 million, whichever is higher.

What does an EU AI Act technical audit involve?

A technical audit under the ACAI framework includes: AI system inventory and Annex III risk classification, calibration testing (ECE), distribution shift detection (PSI), SHAP explainability validation, and adversarial robustness probing. Every finding is mapped to the exact EU AI Act article it violates.

How long does an AI compliance audit take?

The AI Act Readiness Scan takes 2 days and starts from €8,000. The Full Compliance Audit takes 4 weeks and covers all production AI systems across Track A (traditional ML) and Track B (LLM and agentic systems), starting from €25,000.

Do you need access to our production systems?

No. Your team runs all technical tests in your own environment. Gayan de Silva provides the exact protocol, tools and parameters — your engineers execute, and he interprets all results. No external access to production systems is ever required.

ACAI Framework · EU AI Act Audit

Is your AI ready
for EU AI Act
enforcement?

ACAI — AI Compliance Audit & Inspection Framework

A 2-day technical audit that identifies exactly which of your production AI systems are exposed before the August 2026 enforcement deadline. Real engineering tests — calibration testing, distribution shift detection, explainability validation, adversarial robustness probing. Every finding mapped to the exact EU AI Act article.

Check Your Exposure — Free Book a 15-Min Call →

20yr

Production
AI experience

45+

Enterprise
clients

Countries
deployed

€20M

Cost savings
delivered

EU AI Act · Enforcement Deadline

August 2, 2026

High-risk AI system enforcement begins

Days

Hours

Mins

Most DACH companies running AI in production have zero compliance documentation. System inventory takes 2 days. Technical remediation takes weeks. The companies beginning now have time to do this properly.

Check Your Exposure — Free →

Background

IBM·Cisco·4Finance·Zeta Global

Professor of AI · SRH Hamburg

US Patent US12073438B2·GDPR · 17 countries

Free Diagnostic · No Email Required · Instant Results

Do you recognise any
of these five gaps?

The ACAI diagnostic assesses your production AI systems across five technical and governance dimensions. 15 questions. Score out of 75. Per-dimension breakdown with findings mapped to EU AI Act articles.

15 questions 5 dimensions 10 minutes No email required

D1System Inventory & Risk Classification

Annex III · Article 6 · Article 26

Do you know every AI system in production — including vendor tools and embedded models? Have they been classified against Annex III risk tiers? Most organisations discover 30–40% more AI systems than they thought they had.

3 questions

D2Behavioural Consistency & Reliability

Article 15 · Article 61

Have your production models been calibration-tested (ECE)? Are you monitoring for distribution shift (PSI)? Have adversarial inputs been tested? This is the dimension most organisations fail first.

3 questions

D3Data Governance & Lineage

Article 10 · Article 17

Can you trace where your training data came from, how it was processed, and whether it was representative? Undocumented training data is an immediate audit finding under Article 10.

3 questions

D4Governance Process & Change Control

Article 9 · Article 11 · Article 17

Is there a formal deployment gate for AI systems? A model version registry? A continuous risk management process per Article 9? Without these, there is no change control evidence for an enforcement authority.

3 questions

D5Transparency & Explainability

Article 13 · Article 14

Can your AI systems explain their outputs in a way a non-technical decision-maker can act on? Is human oversight for high-risk decisions enforced in practice — not just defined on paper? Article 13 requires meaningful explanations. Article 14 requires genuine human control.

3 questions

What your score means

Critical Exposure

0 – 25 / 75

High Exposure

26 – 40 / 75

Moderate Exposure

41 – 55 / 75

Low Exposure

56 – 65 / 75

Compliant Posture

66 – 75 / 75

Take the Full Diagnostic — Free →

15 questions · Instant results · No email required · Score out of 75 with per-dimension breakdown

The ACAI Audit · Full Technical Assessment

Know exactly where
you stand.
Before August 2026.

A full technical audit of your production AI systems. Every finding mapped to the exact EU AI Act article it violates — with a severity rating and a remediation plan your team can execute.

What you receive

Findings register — every gap mapped to the exact EU AI Act article and severity rating

90-day remediation roadmap — sequenced, with named owners and measurable milestones

Technical test results — real numbers from your production AI systems, not a self-assessment

Board-ready compliance summary — one page your executives can present with confidence

Executive debrief — 60-minute session walking through every finding and the remediation plan

Full Audit Details & Pricing →

What clients and colleagues say

Trusted by people who were there.

★★★★★

Model Governance · GDPR · Zeta Global

Gayan built the model auditing, human-in-the-loop decision frameworks and drift monitoring across our US and European operations years before the EU AI Act made them mandatory. That depth of production experience is exactly what separates a genuine compliance audit from a checklist exercise.

Pawan Pathak

Senior Manager · Zeta Global

★★★★★

GDPR · 17 Countries · 4Finance

Gayan led our GDPR implementation across all 17 country operations — technical controls, data processors and legal alignment. We completed it 18 months ahead of the enforcement deadline with zero compliance issues. The structured approach he brought then is exactly the methodology he applies to EU AI Act compliance today.

Jaroslav Holy

Technology Leader · 4Finance

★★★★★

Data Compliance · EU Customers · Manufacturing

We supply European customers and had no clear picture of our data collection and processing obligations under EU regulation. Gayan mapped our entire data flow, identified where we needed to act and gave us a compliance structure our board could present with confidence.

Tariq Miflal

CEO · Lumala

About

Gayan de Silva, PhD

Professor of AI · SRH Hamburg Enterprise AI Executive PhD · Machine Learning · 2012 EU AI Act · Technical Auditor

Twenty years building production AI/ML systems at IBM, Cisco, 4Finance and Zeta Global. Now Professor of AI at SRH University Hamburg — and the person organisations call when production AI meets an enforcement deadline.

US Patent US12073438B2

45+ enterprise clients · 17 countries

GDPR · 18 months early · 17 countries

Full background and experience →

Is your AI ready
for EU AI Act
enforcement?

Do you recognise any
of these five gaps?

Know exactly where
you stand.
Before August 2026.

Trusted by people who were there.

Insights & Analysis

In 30 minutes, know exactly
where your AI stands.

Is your AI readyfor EU AI Actenforcement?

Do you recognise anyof these five gaps?

Know exactly whereyou stand.Before August 2026.

Trusted by people who were there.

Insights & Analysis

In 30 minutes, know exactlywhere your AI stands.

Is your AI ready
for EU AI Act
enforcement?

Do you recognise any
of these five gaps?

Know exactly where
you stand.
Before August 2026.

In 30 minutes, know exactly
where your AI stands.