If your AI system is high-risk under the AI Act, you can't just deploy it. You need a conformity assessment — proof that the system meets the requirements. Here's what that involves, who can assess, and what happens if you fail.
It's not a self-certification. A conformity assessment is a structured evaluation of your AI system against the AI Act requirements. You must document how your system meets each obligation: risk management, data governance, technical documentation, record-keeping, transparency, human oversight, cybersecurity.
The assessment happens before deployment. Not after. Not "we'll fix it in version 2." You prove conformity, then you deploy.
You can do an internal assessment first, but for high-risk systems, a notified body must be involved. These are organisations designated by EU member states to assess AI systems. The assessment includes:
If the assessment finds gaps, you can't deploy. You fix the gaps, re-assess, then deploy. If you deploy without passing, regulators can pull the system off the market. Non-compliant high-risk AI is illegal to place on the EU market.
The conformity assessment is not a rubber stamp. It's the gate between development and deployment. Skipping it means deploying an illegal product.
Build the evidence during development, not after. If your technical documentation is retrofitted, it will be incomplete. If your risk management was an afterthought, it won't hold up under assessment. Governance architecture must be built in from the start.
Aethel Guard is a Netherlands-based security architecture firm serving EU companies of all sizes. Your architect, not your vendor.