The Human Oversight Specialist makes Article 14 real in DAEIOR: people who can understand, overrule and stop the system, with time, information and authority to do so. Role 1.1 Directs how much autonomy is allowed. Role 5.1 uses the interface. Role 2.3 builds the hooks. This role specifies the oversight model, staffs it, and treats missing hours or a bypassable hook as a defect — not as a training footnote.

Standing duties

• Establish EST07 oversight capacity: roles, competence, roster, deputy, latency budget and authority to stop — matched to the signed purpose and autonomy level.
• Specify Implement hooks with Roles 2.3 and 2.2: override, explanation, queue, fail-closed. A hook that is “available” but not in the path is not oversight.
• Operate live oversight for in-scope systems, or assure the named service owner does, with measured latency and recorded interventions.
• Raise ACT02 when autonomy, speed or population would make the current oversight model inadequate. Raise ACT03 when oversight is bypassed or unmanned.
• Do not treat a training certificate, an “I understand” click or a model card as effective human oversight or as residual-risk acceptance.

Accountable to: AI Governance Lead / Head of Responsible AI (or GM: Operations for live oversight rosters)

Primary domain: Establish EST07 (oversight capacity) and Operate (effective human control). Does not Direct or accept residual risk.

Linked EN 18286 clauses: 4.4.2, 5.3, 5.3.2, 6.1.1, 6.1.2, 6.1.3, 6.2.1, 6.2.2, 8.9.3

Format: 3 hours of training, EUR 200.