Kvíz spojený s průvodcem · Střední úroveň

Kvíz požadavků na lidský dohled u vysoce rizikové AI

Apply Article 14’s proportionate oversight duties, distinguish provider and deployer roles, and identify the limited two-person verification rule.

Otázka 1 z 8

Which purpose does Article 14 assign to human oversight?

Náhled každé otázky8 otázky

Odpovědi a vysvětlení se zobrazují, jak budete v kvízu.

  1. Which purpose does Article 14 assign to human oversight?
    • Prevent or minimize health, safety, or fundamental-rights risks
    • Guarantee every output is correct
    • Transfer accountability to the reviewer
    • Replace risk management
  2. Which factors shape proportionate oversight measures?
    • Model size alone
    • Vendor preference alone
    • Risk, autonomy, and use context
    • Number of users only
  3. What should assigned overseers be able to do under Article 14(4)?
    • Accept every output without review
    • Interpret outputs and decide to disregard, override, or stop the system when appropriate
    • Change training data for each case
    • Guarantee a particular decision
  4. Who identifies provider-side safeguards and integrates them when technically feasible before market placement?
    • The affected person
    • The deployer’s customer
    • A court in every case
    • The provider
  5. When does Article 14(5) require separate verification by at least two people?
    • For every high-risk output
    • For any chatbot response
    • For covered Annex III point 1(a) biometric identification before action, subject to an exception
    • Only for low-risk systems
  6. Does a reviewer’s presence by itself establish effective oversight?
    • No; the person needs relevant information, competence, authority, and ability to intervene
    • Yes, any approval click is sufficient
    • Yes, if the system is accurate on average
    • No, because human oversight is prohibited
  7. How should teams describe the amended AI Act schedule for Annex III systems?
    • Every AI Act duty starts in 2027
    • All AI Act requirements were repealed
    • The change is only a proposal
    • Specified Chapter III high-risk requirements are deferred to 2 December 2027; check scope and transitional rules
  8. Which operational measure can reveal rubber-stamp review?
    • Count only how many people have logins
    • Sample whether staff challenge outputs and review unusually low override rates
    • Assume no complaints proves oversight works
    • Hide override data from managers