Back to blog

Why does EU AI Act compliance still leave builders exposed?

Thắng Đoàn
Thắng Đoàn

You mapped the timeline and maybe certified an AI management system. Then the EU opened its first probes, and customers, boards and regulators kept asking the same question, per fintech.global: who is accountable for AI, and where is the evidence?

A certificate is not compliance

Many firms reach for ISO 42001 as their answer to the AI Act. Per fintech.global, it is a voluntary standard for managing AI through policy, risk processes, named roles and improvement. It does not test models, set accuracy thresholds or approve AI products. Marketing a product as ISO 42001 certified AI overstates what an auditor examined. Think of it as certifying your filing system, not the machine inside it.

  • ISO 42001 governs how you manage AI, not whether your AI is safe
  • It shares structure with ISO 27001, so security teams get a head start

The timeline keeps moving

Compliance is not a fixed target you hit once. Orrick writes on JDSupra that the AI Omnibus modified the AI Act's application timeline with effect from July 27, 2026. Obligations become applicable in stages, so what you owe depends on when you check. A roadmap built last year may already be out of date.

The first probes redraw the line

The Commission's posture is that the research exemption fails the moment an escaped model touches EU-market systems. The EU is investigating OpenAI and Anthropic, Euronews reports, after what the Commission's digital chief described as a string of incidents. This is the first real test of the rulebook's limits, and early cases will set precedents. Per aiweekly.co, the Commission's posture is that the research exemption fails the moment an escaped model touches EU-market systems. Labs that assumed they were out of scope now face obligations.

Governance is the real deliverable

SEV, the Greek business association, frames the AI Act as more than a legal or technical obligation, per euro2day.gr. It calls compliance a matter of strategic readiness and a framework for corporate governance. That means accountability, risk management and clear ownership inside the company. It is the part a certificate alone cannot give you.

What this does not solve

CNN reports Europe is absent from high-level AI safety talks despite writing the first complete rules, and French economist Philippe Aghion warns of an acute risk of marginalization. Being compliant in Europe does not settle where your models fit globally. The precedents from these first probes are still being written, so certainty is not on offer.

If a regulator asked who is accountable for each AI system you run, could you answer with evidence today?
Which of your AI systems do you assume fall outside the AI Act, and what happens if that exemption fails?

List every AI system you operate and name one accountable owner for each; it takes minutes and exposes your real gaps.

Sources

fintech.global - Why ISO 42001 won't make you EU AI Act compliant

JDSupra - The EU AI Act: Key Dates for Compliance

CNN - Europe leads the world on AI rules

euro2day.gr - SEV's new guide for Greek businesses

euronews - EU's AI rules face first enforcement test

aiweekly.co - EU AI Office Opens First AI Act Probes

Share:

Recommended for you

Enjoyed this article?

Subscribe for new articles. No spam. Unsubscribe anytime.

By subscribing you agree to receive the newsletter. No spam, and you can unsubscribe anytime.