AI Act high-risk
A classification under the EU AI Act carrying specific obligations. The decision belongs to the customer's legal and compliance owners, not to engineering.
The EU AI Act sorts AI systems by risk, and the high-risk classification carries the heaviest obligations: risk management, data governance, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness, and cybersecurity.
Why engineering cannot decide it
Classification turns on intended purpose, the role the organization plays, the domain of use, and legal interpretation of the text. Those are legal determinations. An engineering team can establish the facts a classification depends on — what the system does, who it affects, what it decides, what evidence exists — but stating the facts is not the same as reaching the conclusion.
What engineering can document
A fixed system boundary and intended-purpose record. A log of applicability questions with working hypotheses, the facts that support them, what remains unresolved, and which named owner decides each one. A control–evidence matrix mapping conditionally relevant requirements to inspectable policies, tests, logs, runbooks, and records — including where evidence does not yet exist.
What it is not
Not legal advice. Not classification. Not certification, conformity assessment, or CE marking. Not a compliance guarantee, and not continuing assurance. The record is dated, limited to the configuration assessed, and reopened when purpose, actors, data, model, tools, deployment, incidents, or the law change.
Often confused with
How it differs from similar terms.
| AI Act high-risk | Compared with | What's different |
|---|---|---|
| Engineering record | Conformity assessment | An engineering record prepares evidence and names open questions. Conformity assessment is a formal legal procedure eqlx does not perform. |
Want these terms worked through on one of your own decisions? Start with a Decision Blueprint