To let natural persons effectively monitor the system: What is the primary purpose of this requirement?

Human oversight lets a person monitor, intervene in, or stop a high-risk system to prevent or minimize risks to safety and rights.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. To ensure a human manually re-enters and retypes every input into the model, so the automated system never processes any data that a person has not first transcribed.

   Plausible because oversight involves people, but the goal is meaningful monitoring and intervention capability, not manual re-entry of all inputs.
2. To guarantee that the deployed model reaches perfect predictive accuracy, since a human reviewer is expected to correct every output before it is ever shown to any affected person.

   Almost right that reviewers add value, but oversight aims at risk control and intervention, not at guaranteeing perfect accuracy on every output.
3. To transfer all legal responsibility for the system onto the individual human overseer, relieving the provider and the deployer of their own statutory duties under the Act.

   Plausible-sounding but oversight does not shift statutory duties onto one person; providers and deployers retain their respective obligations.
4. To let natural persons effectively monitor the system and intervene or stop it, so risks to health, safety, or fundamental rights can be prevented or minimized during use. ✓

   Correct: human oversight ensures a person can understand, monitor, and if needed override or halt a high-risk system to keep risks in check.

How to remember it

Human oversight lets a person monitor, intervene in, or stop a high-risk system to prevent or minimize risks to safety and rights.

More Understanding How Laws, Standards and Frameworks Apply to AI questions

• Establish a risk-management system and maintain technical: Under the EU AI Act, which requirement is a core →
• Maintain technical documentation and provide information: Under the EU AI Act, which obligation applies to →
• It uses tiered maximums scaled to the violation's: How does the EU AI Act structure its administrative fines →
• All 394 Understanding How Laws, Standards and Frameworks Apply to AI questions →