Page 371 - CyberSecurityEssentialsEnglish

TABLE 8-5: A SIMPLIFIED INSTITUTIONAL CLASSIFICATION OF USES

Level Typical examples What is required

Rewording an internal note, Standing rules and user review
Low
brainstorming, formatting. before use.

An internal summary, a first analysis, A registered owner, an acceptance
Medium
a knowledge assistant. test, and structured review.

Ranking applicants, prioritizing a An impact assessment, expert

High service, a health or legal review, documented approval, and
recommendation. monitoring.

Uses that breach law, rights, or Do not use; escalate for a
Prohibited
institutional policy. leadership decision.

● Conditions for meaningful human oversight

• Competence: the reviewer understands the task, the system, and its limits.
• Information and time: they see the sources and assumptions and have real time to
check.
• Authority and independence: they can reject, amend, or escalate without penalty.
• Accountability and appeal: the decision is documented and the affected person can
request human review.

Video: AI Literacy and Prohibited Practices in the
European AI Act
A webinar on the relationship between training and

risk of use, and what is expected of organizations
and of the people who operate the systems.
Duration: about 60 minutes | Level: Intermediate

Learning objective: Build a graduated literacy
program by role, context, and risk instead of
uniform training.

Source: European Commission [18].
https://www.youtube.com/watch?v=8gtaiZhAVNQ
Verified: June 6, 2026

The Right to Object and Procedural Fairness
When a system affects a person, good average performance is not enough. Procedural
fairness must be considered: does the person know that AI took part in the process? Can

they understand the main factors that shaped the outcome? Is there a channel for correcting

370