Ikibazo gihuza ibibazo · Hagati Urwego

Amategeko y’ubumwe bw’ibihugu by’Uburayi agenga amanota yo gutanga inguzanyo hamwe n’ikibazo cy’ibiciro by’ubwishingizi

Gerageza iki gitabo cy’ubumwe bw’ibihugu by’Uburayi AI, inshingano, inshingano, imipaka ifatika, hamwe n’ibitekerezo byoherejwe ku mashyirahamwe.

Ikibazo 1 Bya 8

Which credit use is named as high-risk in Annex III?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Which credit use is named as high-risk in Annex III?
    • Evaluating a natural person’s creditworthiness or credit score
    • Detecting financial fraud
    • Sending a payment receipt
    • Forecasting a bank’s office expenses
  2. Which insurance use does Annex III specifically identify?
    • Every form of insurance pricing
    • Property maintenance scheduling
    • Risk assessment and pricing for natural persons in life or health insurance
    • Reinsurance reserving in every circumstance
  3. When do Annex III requirements apply under the current timeline?
    • 2 February 2025
    • 2 Ukuboza 2027
    • 2 Kanama 2026
    • 2 Kanama 2028
  4. Does the fraud-detection exception mean all fraud tools are outside the AI Act?
    • Yes, all financial AI is excluded
    • Yes, if a bank deploys it
    • No, because fraud detection is always prohibited
    • No; it is an exception to this specific Annex III creditworthiness entry
  5. What pipeline elements should a firm trace when auditing a score?
    • Only the model architecture
    • Gusa amanota yanyuma
    • Data, features, score, threshold, human exception, and customer communication
    • Only the vendor’s marketing claim
  6. Does high-risk classification establish a lawful basis for personal data?
    • No; data protection law still requires a separate assessment
    • Yes, for all financial institutions
    • Yes, if the model is calibrated
    • Only when the data are public
  7. Why monitor drift after deployment?
    • Drift proves fraud
    • It removes the need for logs
    • It automatically fixes biased data
    • Changing products or populations can alter model performance and harms
  8. What should a firm reassess if a model’s purpose changes?
    • Gusa izina ry'icyitegererezo
    • Whether its AI Act classification and other legal obligations change
    • Nothing if the code is unchanged
    • Only its cloud provider